How we pick traders Hyperliquid · counts read 07 Nov 2026 19:48 UTC
We are building a short list of Hyperliquid traders worth following. Each must profit over many trades, not one lucky bet, hold long enough to follow, take survivable risk, and still be trading. Then we will test whether the list predicts the next 90 days. Research only: the site never places a trade.
Runs on its own
Census and profit floor every 48 hours, the screen every day: 1,264 wallets passed and are shown on the Dashboard, Tracker and Open Interest. Their positions are re-read in turn, up to 950 reads every 30 minutes.
Built, run by hand
The deep checks, steps 4–10: 108 wallets checked so far, 0 cleared. They are not on a schedule yet.
Not started
The forward test, step 11. No list has yet been checked against the 90 days that followed.
Blocked by history
Hyperliquid gives a busy account only its last 50–80 days of trades, so for a two-year trader the trade-by-trade checks (steps 6–7 and parts of 8 and 10) see only its last two months. What could fix it.
The steps each box is a set of yes/no checks; a wallet goes on only if it passes all of them
runs on its own
only when started by hand
not running
Numbers are wallets on file today. A fill is one executed piece of an order; one
trade is often many fills.
- Runs on its own
- Data inHyperliquid leaderboardevery address on it1
Census
runs on its own · every 48 hoursWho can we see?
A wallet that never shows up on the leaderboard is never screened automatically; it can only be added by hand.
All 9 rules for this step, with exact values - 51,190 wallets on file
- Data inHyperliquid leaderboardall-time profit from all its trading2
Profit floor
runs on its own · every 48 hoursHas it made serious money?
- Made $500K or more over its life?
no →Under $500K lifetime 46,325
- 4,334 made $500K or more
- Data inHyperliquid account dataprofit split and account valueHyperliquid account datastaking rewards, at today’s HYPE priceHyperliquid account datatrades in the last 90 days3
Daily screen
runs on its own · every day, most profitable firstStill trading, with real money?
- Account worth $1M or more?
no →Account under $1M 2,433- Traded in the last 90 days?
no →Quiet 90 days 832
It also names where the profit comes from: perps (futures), spot coins or staking. That is a label, not a removal. A wallet that fails is screened again after 30 days; one that passes is not screened again.
All 29 rules for this step, with exact values - 1,065 passed, plus 199 that passed and have since fallen under $500K: 1,264 shown on the Dashboard, Tracker and Open Interest4 not screened yet: the daily screen has not reached them
- Built, but runs only when started by hand · 108 of the 1,264 that passed checked so far
- Data inHyperliquid account dataowner, sub-accounts, positions, coins, stakingHypurrscanpublic address namesSaved listsvaults and builder exchanges, copies from 11 Sep 20264
Who is behind it
only when started by handIs this one person betting on price?
- Belongs to another account (a sub-account, or a trading key it authorised)?
yes →Sub-account or Agent wallet 0- A pooled fund (vault), or its manager?
yes →Is a vault or Runs a vault 6- A known exchange, bridge or protocol address?
yes →Exchange or protocol 0- Half its money or more in staked HYPE?
yes →Staker 29- Holds coins that cover half or more of a bet against the same coin?
yes →Basis / hedged 12
These are checked before any trade history is downloaded; four more “who is behind it” checks need that history, so they come in step 5. If the owning account of a sub-account is not on file yet, it is queued. Only wallets that passed the daily screen are taken, plus wallets added by hand; a wallet the screen failed goes back to the screen.
All 21 rules for this step, with exact values - Data inHyperliquid account dataits 2,000 latest fills first; then its fill history, deposits, transfers, fundingHyperliquid live feedthe trade recorder (switched off)Hyperliquid fills archiveplanned, not connected: older history5
Rebuild every trade
only when started by handWhat did it actually do?
- Its download finished?
no →not judged; tried again next run- Quick look at its latest 2,000 fills: 30+ exits held under 5 minutes, making 30%+ of its winning profit?
yes →Scalper 0- Its fills add up, with no gaps?
no →not judged; tried again next run- Too many fills to rebuild: 120,000+ even from a later start?
yes →History too deep to rebuild 1- More than 80% of its fills come from orders it left waiting?
yes →Market maker 10- Funding is more than half its profit?
yes →Carry desk 2- Under $5 profit for every $10,000 traded?
yes →Volume farmer 2- Ever liquidated?
yes →Has been liquidated 3- A holder: under 30 trades with a typical hold of a week or more, or no closed trade in 30+ days?
yes →Long-term list: skips steps 6–10, never one to follow 4- Trades against wallets it is linked to?
yes →Wash trading 0- Copies another wallet’s trades within 2 seconds, again and again?
yes →Copies someone else 0
Fills are grouped into trades that run from no position back to no position. A trade already open when the history starts is not counted. The wash and copy search is a separate run that mostly needs the trade recorder, which is off; today it can only compare wallets whose trades we have stored.
All 39 rules for this step, with exact values - Data inRebuilt tradesfrom step 56
One lucky bet?
only when started by handIs the profit spread over many trades?
- 50 or more closed trades?
- Best 3 trades at most 40% of the profit?
- Profit spread like 15 or more equal winners?
- Still $500K or more without its best trade?
- And without its best 30 days?
fails any →Failed lucky-bet test 39
- Data inRebuilt tradeshold times7
Scalper?
only when started by handDoes it hold long enough to follow?
- At most 10% of trades closed within a minute?
- At most 30% within 5 minutes?
- A typical hold of an hour or more, per trade and by size?
fails any →Failed hold-time test 0
Bots are fine if they hold for hours: “bot” is only a label.
All 12 rules for this step, with exact values - Data inRebuilt tradessingle-trade lossesAccount-value historyits whole life, for the worst fall8
Risk
only when started by handCould it blow up?
- Worst fall from a peak 35% or less?
- No trade lost more than 15% of the account?
- Never liquidated?
fails any →Failed risk test 0
- Data inAccount-value historyand the span of its trades9
Track record
only when started by handLong enough to judge?
- Under 90 days of record?
yes →Too young 0
90 to 270 days: a candidate, a watchlist only. 270 days or more: proven, and on to step 10.
All 3 rules for this step, with exact values - Data inRebuilt tradesby month, when the history is completeAccount-value history30-day returns, first half against second; months too when the trades are cut shortSaved copiesBitcoin daily prices and the 40 busiest markets, 11 Sep 202610
Consistency
only when started by hand · proven traders onlyA steady edge, or one good season?
- Profitable in 60% of months or more?
- No month over 40% of the total?
- 30-day return positive 55% of the time or more?
- Second half at least half as good as the first?
- Worst loss at most 10× its typical loss?
- Size not changed after wins or losses?
- Makes money on the side it trades less (20+ trades there), or, if it trades almost only one way, when Bitcoin’s trend ran against that side?
- A setup it repeats that makes money?
fails any →Failed consistency test 0
Listed traders are checked again only when someone starts a re-check by hand. Failing any check drops them; a weak last 60 days marks them “Slipping”, and two in a row drops them.
All 25 rules for this step, with exact values - 0 cleared every check (candidates and proven)
- Not started
- Data inThe frozen listand 50 random wallets that failed step 6, 7, 8 or 1011
Forward test
not runningDoes the list predict anything?
Freeze the list, then follow it, including a copy that enters 2 seconds late. The aim is 90 days, but tracking has no end date yet. A weak link between rank and the next months’ return gives “do not use it”; a clearly backwards one does not yet, a known gap. The checks are never loosened to make names pass.
All 15 rules for this step, with exact values
Where the data comes from every source the steps and the pages use, what we ask it for, how often
- Hyperliquid leaderboard
stats-data.hyperliquid.xyzEvery address on the board, with its all-time profit, account value and volumeevery 48 hoursSteps 1–2 - Hyperliquid account data
api.hyperliquid.xyz/infoFor each $500K+ wallet not yet screened: its profit split, account value, staking rewards and trades in the last 90 days; and the HYPE priceevery day, up to 900 look-upsStep 3 - Hyperliquid account data
api.hyperliquid.xyz/infoOpen positions and spot balances of every wallet that passed the screen; each day’s last reading of the positions is kept 8 days for the 3D and 7D columnsevery 30 minutes, up to 950 reads a run (the rest go first next run)Dashboard, Tracker, Open Interest, trader pages - Hyperliquid account data
api.hyperliquid.xyz/infoWhen each open position and holding began (from recent trades), and futures-only profitevery 3 hours to dailyDashboard, Tracker, Open Interest - Hyperliquid account data
api.hyperliquid.xyz/infoThe account-value and profit history of each newly passed wallet, read onceevery 3 hoursTrader pages - Hyperliquid account data
api.hyperliquid.xyz/infoEvery trade of each wallet that passed the screen, as far back as Hyperliquid still serves it, rebuilt into round trips (in and out, size, profit) and kept; then the trades sinceevery 30 minutes, up to 300 look-ups a run (a wallet every 6 hours if it traded this week, else daily)Trader pages - Hyperliquid live feed
api.hyperliquid.xyz/wsLive trades of 15 watched wallets, and live pricesalways onDashboard, Open Interest - Hyperliquid account data
api.hyperliquid.xyz/infoIts owner, positions, spot and staking first (a wallet removed at step 4 stops there); then its 2,000 latest fills; then its fill history (everything the API still serves, up to about the latest 100,000), deposits, transfers, funding over the same span, and account history20 wallets a runSteps 4–10 - Hypurrscan
api.hypurrscan.ioPublic names and tags of exchange, bridge, vault and protocol addresseseach run, kept a dayStep 4 - Saved copies
downloaded once, 11 Sep 2026Hyperliquid’s vault list, its builder exchanges, its 40 busiest markets, and Bitcoin daily pricesnot refreshedSteps 4 and 10 - Hyperliquid live feed
api.hyperliquid.xyz/wsEvery trade in the 40 busiest markets, with buyer and sellerrecorder switched offStep 5 (wash and copy search) - Aster, Pacifica, Lighter
each venue’s own APIEach venue’s leaderboard, the open positions of its top traders (each day’s last reading kept 8 days, as on Hyperliquid), trade history for their pages, live priceshourly to every 2 hoursDashboard, Tracker, Open Interest, trader pages - Hourly open interest
the position reads aboveNo extra request: every read of an account’s open positions, on all four venues, is also kept by the hour (the hour’s last reading) for the 7-day open interest lines and the OI High/Low and Net OI High/Low columns, kept 8 daysHyperliquid every 30 minutes, Lighter and Pacifica hourly (the board’s top 40), Aster every 2 hoursDashboard, Tracker, Open Interest, trader pages
Almost everything we read is stored in our own database, and pages are drawn from there; only Aster, Pacifica and Lighter prices are fetched live while a page is built, and kept for a minute. A wallet that fails a check is never deleted: it stays on file with its reason. Only raw logs age out: recorded market trades after 14 days, position changes after six months. Aster, Pacifica and Lighter traders are listed from each venue’s own leaderboard; none of the steps run on them.
Next: more history what could fix the short history
Hyperliquid gives a busy account (over about 10,000 fills) only its
last 50–80 days of trades. It also publishes fills on Amazon’s cloud,
written by one of its own nodes
(s3://hl-mainnet-node-data/node_fills_by_block). If we can use them,
steps 5–10 get older history (how far back is question 2). We could also run the
forward test on the past: pick the list as of an old date, then check the next 90 days.
Testing one day of it answers four questions. None is answered yet: Amazon
charges whoever downloads, and we have not set that up.
- 1
Does every fill carry the wallet address?
Today we fetch one wallet’s trades at a time, so we know whose they are. The archive is a node’s record of fills, one block per line, and its documentation does not say if each fill names its wallet. Our live trade feed names buyer and seller: a good sign, not proof.
Test: read one file - 2
How far back does it go, including the older format?
The documentation names three folders, the current one and two older ones in other formats, with no dates for any.
Test: list the folders - 3
How big is one day, and what would the full history cost?
Unknown until the files are listed. Whoever downloads pays Amazon’s request and transfer fees: roughly $0.09–0.12 a GB over the internet, almost nothing to a server in the same Amazon region. Keeping a copy costs our own storage.
Test: list one day’s files - 4
Do one wallet’s archive fills match the API for the same day?
This decides if the archive is usable, and whether it holds every fill. Every wallet we rebuild must add up: each fill starts from the position the one before it left. The same test on archive fills, for a wallet we already hold, settles it.
Test: compare one wallet
Who we leave out the six kinds that need explaining; the steps also remove others, such as stakers, vaults and quiet accounts
One-bet whale
Made most of its money on one or two huge bets. Take those away and little is left.
How we spot it todayStep 6, run by hand.
More detail
An account whose profit comes from one or a few oversized bets, or one lucky stretch. Take that away and little or nothing is left.
What it looks like
- The profit curve is a single step, flat before and after
- The best trade or best month is half the profit or more
- One coin, one direction, a size far above its usual
- The big bet is often still open, or began before the history we can see
Our test today
Step 6, run only when the deep check is started by hand (the daily screen does not apply it): at least 50 closed trades, the best 3 trades at most 40% of gross profit, at least 15 ‘effective’ winners, and still $500K+ after removing the best trade and the best 30-day block.
Where it falls short- The 37 wallets removed this way have a median of 2 closed trades in about 11 days of visible history — too little to judge, not one lucky bet
- Busy accounts only show 50–80 days, so the $500K left after dropping the best 30-day block has to come from the other 20–50 days: roughly a $4–9M-a-year pace
- It misses a real whale that scales one idea in and out 60 times, or whose big bet is still open
Better definition · planned as a label, not built yet
- First: do we see enough? If the fills we can read cover under 90% of the account’s traded volume, it is ‘not enough history’ — neither a pass nor a whale
- Evidence: a profit per trade unlikely to be luck (a standard statistical test), also on weekly totals so one idea split into many trades counts once
- Whale: the best trade is 50%+ of profit and removing it leaves nothing; or, with 270+ days of history, the best 30 days is 60%+ of lifetime perp profit and removing it leaves nothing
Scalper and high-frequency bot
Holds a trade for seconds or minutes, often a bot. Too fast to follow: a copy arrives after the move.
How we spot it todayStep 5 (a quick look at its latest fills) and step 7 (the history we download), run by hand.
More detail
A scalper holds for seconds to minutes and earns a few dollars per $10,000 traded, which a copy delay and fees wipe out. A high-frequency bot is the machine version: 24/7, hundreds of orders a day, inventory near flat. A bot that holds for hours is fine.
What it looks like
- Each unit of size held for minutes, not hours
- Dozens of round trips a day; daily volume many times the account
- Trades in every hour of the day
Our test today
Step 7, run only when the deep check is started by hand: at most 10% of trades under a minute, a typical hold of at least an hour (per trade and weighted by size), at most 30% under 5 minutes. Separately (step 5), before a wallet’s first full download its 2,000 most recent fills are read: 30 or more exit orders held under 5 minutes that carry at least 30% of the timed winning profit remove it as a scalper, never re-checked. Otherwise the download starts late enough to hold about 100,000 fills, and only an account that still reaches about 120,000 is removed as ‘too deep to rebuild’. It is due another look after 60 days, but only when the re-check is run by hand.
Where it falls short- Scalping around a core position passes: the hold is measured on the core
- Step 7 counts trades, not money: 2-minute trades that make 98% of the profit pass if they are a quarter of the trades (the quick look at the latest 2,000 fills does weigh profit, but only on those fills)
- It removes good traders: a breakout trader with tight stops, a 45-minute trader
Better definition · planned as a label, not built yet
- Match every exit to the size it closes across the whole history, not only the latest 2,000 fills; a scalper makes more than 30% of its profit on exits held under 5 minutes, with 30 or more of them
- Between 5 minutes and an hour, only an edge under $15 per $10,000 traded counts against it
- Profiles of checked traders already show a style by typical hold (scalp, intraday, swing, position); the plan adds high-frequency and market maker
Wash trading
One owner trading with itself through linked wallets, to fake volume or move profit between them.
How we spot it todayStep 5, run by hand. It finds nothing today: the trade recorder it needs is off.
More detail
One owner on both sides of trades through linked wallets, to fake volume or rank, or to move profit from one wallet to another. Hyperliquid blocks a wallet trading with itself, so it takes two or more.
What it looks like
- The same two wallets on opposite sides of the same trade, far more often than their share of volume predicts
- Linked by a transfer, a sub-account or a common funder
- Trades that undo each other within minutes, so the pair stays flat while its volume grows; one wallet’s gains are the other’s losses
Our test today
Step 5, by hand only: a graph of linked wallets plus buyer/seller pairs from the live trade feed — at least 20 trades together, and at least 10% of a wallet’s fills against the other; every wallet in a group whose members trade against each other is removed for good.
Where it falls short- It runs only by hand, and the trade-feed recorder it needs is off, so it finds nothing today
- Two market makers quoting the same thin book get a trading-against-each-other link (it removes them only with a second signal or a transfer); chains of links reach unrelated wallets; a single transfer joins wallets into one group, and if any two members trade against each other often enough, every member is removed for good
Better definition · planned as a label, not built yet
- A real link first: a sub-account, a direct transfer, or a shared funder that is not an exchange
- Add up the linked wallets’ profit and judge them as one trader
- Wash when the dollars traded between them are 10× or more their net change, half reversed within an hour; an unlinked pair is only flagged for review
Copy-follower
Repeats another wallet’s trades a moment later. Following it adds nothing: follow the original.
How we spot it todayStep 5, run by hand. The trade recorder it mostly needs is off, so it can only compare wallets whose trades we have stored.
More detail
A wallet that replays another wallet’s trades with a delay. Following it adds nothing: follow the leader instead.
What it looks like
- Same coin, same direction, after a tight, repeatable delay (under 3 seconds for a bot, minutes for a person)
- It mirrors the exits too, sizes in proportion, and almost never goes first
- No funding link to the wallet it copies
Our test today
Step 5, by hand only: liquidity-taking fills in the same coin and direction within 2 seconds after another wallet’s, at least 20 times, on at least 30% of its liquidity-taking fills, and at least 5× more often than chance — not when the two are linked by a transfer or sub-account, or echo each other both ways.
Where it falls short- It runs only by hand, and the trade-feed recorder it mainly reads is off, so today it can only compare wallets whose fills the deep check has stored
- It counts fills, not decisions: one order that sweeps 25 price levels counts as 25 copies
- News moments get the slower of two independent traders removed; it misses bots 5 seconds behind and copiers of several leaders
Better definition · planned as a label, not built yet
- Count entry decisions within a delay of up to 5 minutes, and require the delay to be steady
- At least 20 matches over 5 or more days, no more than half on one day, 40% or more of its entries, exits mirrored
- It has to beat chance against similar traders on the same coin and side
Hedger
Makes money from offsetting positions or funding payments, not from being right about price.
How we spot it todayStep 4 (coins held against a bet on the same coin) and step 5 (funding), run by hand.
More detail
Profit that does not come from price direction: long spot or staked HYPE against a short perp, sitting on the side that is paid funding, a pairs book, or legs split across sub-accounts or exchanges.
What it looks like
- Opposite legs in the same coin
- Net exposure small next to gross exposure
- Funding received far above funding paid; profit that does not move with the market
Our test today
Steps 4 and 5, run only when the deep check is started by hand: a perp short at least 50% covered by the same coin held as spot, at the moment of the check (step 4); net funding, paid or received, more than half of the profit or loss over the trades we downloaded, only the last 30 days when a wallet is checked again (‘carry desk’, step 5). A low-volatility test exists but is switched off.
Where it falls short- It is one snapshot
- A small hedge removes a directional account: a $148K HYPE short against a HYPE holding removed a wallet whose $7.2M of ZEC and XMR shorts were plainly directional
- It misses hedges with staked HYPE, in sub-accounts or on other exchanges; ‘carry’ also counts funding paid, not only received
Better definition · planned as a label, not built yet
- Per coin, how much of the perp position is offset, spot and staked HYPE included
- Basis: 50% or more offset on at least 20 of the last 30 daily readings
- Carry: funding received is 50%+ of perp profit and a hedge is present; a pairs book (net under 25% of gross for 60+ days) is a label. A hedge on a centralised exchange cannot be seen from Hyperliquid
Market maker
Posts buy and sell orders all day and earns the small gap between them. You cannot copy a quote.
How we spot it todayStep 5, run by hand.
More detail
Quotes both sides, earns the spread and rebates, gets back to flat fast, and keeps about $0.50–3 per $10,000 traded. You cannot copy a quote.
What it looks like
- Mostly passive fills, measured in dollars
- A passive buy and a passive sell on the same coin within a minute, again and again
- Inventory swinging around flat; negative fees (rebates) on passive fills
Our test today
Step 5, run only when the deep check is started by hand: more than 80% of fills are passive (perp fills, counted by number of fills). A second test removes a ‘volume farmer’: all-time profit under $5 per $10,000 of all-time volume (leaderboard figures, spot included).
Where it falls short- The 8 wallets removed as market makers hold for a median of about 25 hours and keep about $72 per $10,000 traded — patient limit-order traders: one resting order filled in 40 pieces counts as 40 passive fills
- The 2 wallets removed as ‘volume farmers’ look like the real market makers: billions traded at $2–5 per $10,000, holding 11 minutes
Better definition · planned as a label, not built yet
- Passive by dollars: 60%+ of the dollars traded (split-up orders left out)
- Two-sided: 30%+ of that passive money has an opposite passive fill on the same coin within 60 seconds
- And a thin edge (under $5 per $10,000) or fast inventory turnover; rebates on passive fills are proof on their own. Anything else is labelled a ‘limit-order trader’
Every rule, with its exact value 402 rules · 216 run
only by hand · 13 written but not in force
The full list behind the steps: what each rule does, the value it uses (for rules marked •, read from the code when this page is drawn), whether it removes a wallet or only labels it, how often it really runs, and whether the value is a starting value or a decision. 194 values marked • were compared with the code at 07 Nov 2026 19:48 UTC and agree with it.
402 rules shown
1 · Census Who can be seen at all. · 9 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Where wallets come fromWallets enter the census only from Hyperliquid's public leaderboard. An address that never appears there is never screened automatically.Other ways in (the live trade tape, traders found being copied) exist but only run by hand. | — | Routes | Every 48 h | Starting value |
| Census refresh every 48 hoursEvery 48 hours the leaderboard is downloaded again and every wallet on it is re-judged against the $500K profit floor, so new wallets can enter.Deeper verdicts and the daily screen's own removals are kept. A wallet that passed the daily screen but has not had its deep check is re-judged like any other: under $500K now, it is set aside for the profit floor and leaves the deep-check queue, though it stays on the site. | 48 h | Schedule | Every 48 h | Starting value |
| Reuse a fresh downloadIf the saved leaderboard file is less than 36 hours old, the refresh reuses it instead of downloading it again.In practice this matters when a failed refresh is retried 15 minutes later. | 36 h | Schedule | Every 48 h | Starting value |
| Reject a truncated downloadA leaderboard download smaller than 5 MB is treated as broken and thrown away. The previous census stays in use, and the download is tried again 15 minutes later.A normal download is about 37 MB. | 5 MB | Schedule | Every 48 h | Starting value |
| Deeper verdicts left aloneA census refresh leaves alone every wallet with a deeper verdict: candidate, proven, slipping, too-young and long-term wallets, and wallets removed by a later check.The leaderboard figures on file for these wallets, such as lifetime profit, are no longer refreshed. | — | Routes | Every 48 h | Fixed |
| Trade tape: one big fill Runs by hand onlyWhen the live trade-tape listener is running, any address that makes a single fill of $25,000 or more on the 40 busiest perp markets is added to the census as a new wallet.The listener is not part of the automatic setup, and wallets it adds are screened only by a separate by-hand run. | $25,000 • | Routes | By hand | Starting value first guess |
| Trade tape: $500K in a week Runs by hand onlyWhen the listener is running, any address that trades $500,000 or more in total over a rolling 7 days on those markets is also added as a new wallet.Only trades seen while the listener is running count toward the 7 days. | $500,000 • | Routes | By hand | Starting value first guess |
| Trade tape: 40 busiest markets Runs by hand onlyThe trade-tape listener follows the 40 perp markets with the most daily volume, picked when it starts.Not part of the automatic setup. | 40 | Routes | By hand | Starting value |
| Copied traders enter the census Runs by hand onlyWhen a wallet is found copying another wallet's trades, the wallet being copied is added to the census as a new wallet, if it is not already on file.Copy detection runs only by hand, and a wallet added this way is screened only by a by-hand run unless it later appears on the leaderboard. | — | Routes | By hand (after a manual review step) | Starting value |
2 · Profit floor The lifetime-profit bar every wallet has to clear. · 2 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Profit floorA wallet needs at least $500,000 of all-time profit on the Hyperliquid leaderboard, perps and spot together. Wallets below that are set aside with that reason and kept on file.This is checked again at every census refresh, and a wallet that has already passed the screen stays on the site if it later slips below. | $500,000 • | Removes | Every 48 h | Owner's decision |
| Off-leaderboard wallets: $500K perp profit Runs by hand onlyA wallet found outside the leaderboard (on the live trade tape, or as a copied trader) needs at least $500,000 of all-time perp profit, read from its own account history.Only runs by hand, and it counts perp profit alone, while the leaderboard floor counts perps and spot together. | $500,000 • | Removes | By hand | Starting value |
3 · The screen Account size, recent activity and where the profit comes from. · 29 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Who the daily screen picks upEach day the screen takes every wallet on file with at least $500K of lifetime profit that it has not screened yet. It also takes back any wallet it removed itself (small account, no recent trade) once that wallet's 30-day re-check date has passed, and, on the same 30-day clock, any wallet that failed it while holding a deeper verdict or after being put back in the queue. A typed wallet waiting in the queue without a pass, which failed the screen and lost its reason, is taken at once.A wallet that passed is not screened again, so its pass stands even if the account later shrinks or goes quiet; the top-trader lists drop a passer with no trade in 30 days unless it holds a running trade, and the 'Traded within' filter drops a quiet one; both count from its last trade's date. | $500,000 • | Schedule | Daily | Owner's decision |
| Most profitable firstWallets are screened in order of lifetime profit, largest first, so if the day's budget runs out the most profitable wallets have already been done.New wallets and due re-checks share the same queue. | — | Schedule | Daily | Owner's decision |
| Daily request budgetThe daily screen stops after 900 requests to the exchange. Wallets it did not reach wait for the next day's run.A new wallet or an account-size re-check costs up to 4 requests (portfolio, staking rewards, and one or two for its fills), or 1 if its account is under $1M; a quiet wallet's re-check costs at most 2. A trade on record in the last 90 days saves the fills requests. | 900 | Schedule | Daily | Owner's decision |
| Order the rules run inAfter the profit floor, the screen checks account size first, then splits profit into perp, spot and staking, then checks for a recent trade. A wallet stops at the first rule it fails.A quiet wallet's 30-day re-check runs only the recent-trade rule. | — | Schedule | Daily | Owner's decision |
| $1M accountThe account must be worth at least $1,000,000 by the exchange's own figure, read from its portfolio every time the wallet is screened or re-checked. This also removes traders who have withdrawn their winnings.If the portfolio has no figure, the latest leaderboard figure is used instead; the census re-reads it every 48 hours. An older stored figure is never used. A wallet that passed is not screened again, so it keeps its pass if the account shrinks later. | $1,000,000 • | Removes | Daily | Owner's decision |
| Confirm before removing on sizeIf the leaderboard puts the account under $1M, the screen reads the account's own history once more and uses that figure before removing it.This is skipped when an older account reading is already on file, and that reading is used instead. | — | Removes | Daily | Fixed |
| Staking profit at the day's HYPE priceStaking profit is every HYPE staking reward the exchange lists for the wallet, added up and valued at the HYPE price on the day of the screen.It is not revalued later, and if no HYPE price is available that day it counts as zero. | — | Labels | Daily | Owner's decision |
| Spot profit is the remainderPerp profit is the exchange's own perp-only lifetime figure. Spot profit is whatever is left of total lifetime profit after perps.Both figures are read once, sometimes at an earlier check, and spot profit also takes in any other non-perp profit. | — | Labels | Daily | Owner's decision |
| Perp trader, spot trader or stakerWhichever of perp profit, spot profit and staking profit is largest decides the wallet's type: perp trader, spot trader or staker. All three types pass, so this split never removes anyone.Signed figures are compared, so a wallet with a perp loss and spot gains is a spot trader. | — | Labels | Daily | Owner's decision |
| Traded in the last 90 daysThe wallet must have traded in the last 90 days, otherwise it is removed. A trade on record inside those 90 days, counted from its date to today, is enough on its own; otherwise the wallet's fills for the 90 days are read, and the newest one becomes its last-trade date.The fills the screen reads leave out TWAP order slices, but a date recorded by another job counts whatever trade it came from. An old yes-or-no answer with no date, such as the 12 Sep 2026 probe left, is never reused. A removed wallet is checked again after 30 days. | 90 days • | Removes | Daily | Owner's decision |
| What passing doesA wallet that passes is marked as passed and gets its type. A wallet holding a deeper verdict (candidate, proven, slipping, too young, long-term, or removed by a later check) keeps its state, its reason and its re-check date and only gets its figures updated; any other wallet joins the queue for the deeper checks.A wallet that fails the size or 90-day rule while holding a deeper verdict keeps it too; the failure is noted on its record and checked again after 30 days. Two retired verdicts of the 12 Sep 2026 one-off runs give way: a removal for low return is replaced by the screen's own result, and a spot holder those runs put on the long-term list joins the deep-check queue if it passes, or stays on the list if it fails. | — | Routes | Daily | Fixed |
| Optional pass: no perp history Runs by hand onlyWhen the optional portfolio pass is run, a wallet whose perp history for the last month or the last week cannot be read is removed.Only runs by hand, and the next census refresh re-judges any wallet it removed. | — | Removes | By hand | Starting value |
| Old full screen: 50% return Runs by hand onlyIn the old full screen, which now runs only if someone starts it by hand, the leaderboard's all-time return had to be at least 50%.The leaderboard's return is not adjusted for deposits, and the next census refresh re-judges anyone this removed. | 50% • | Removes | By hand | Starting value |
| Old full screen: $1M account Runs by hand onlyIn the old full screen only, the leaderboard's account value had to be at least $1,000,000.This is not the live $1M rule, and the next census refresh re-judges anyone this removed. | $1,000,000 • | Removes | By hand | Starting value |
| Old full screen: green month Runs by hand onlyIn the old full screen only, the leaderboard's profit for the last month had to be above zero.Not part of the live screen, and the next census refresh re-judges anyone this removed. | on • | Removes | By hand | Starting value |
| Old full screen: green week Runs by hand onlyIn the old full screen only, the leaderboard's profit for the last week had to be above zero.Not part of the live screen, and the next census refresh re-judges anyone this removed. | on • | Removes | By hand | Starting value |
| Old full screen: 5 bps edge Runs by hand onlyIn the old full screen only, lifetime profit had to be at least 5 basis points (0.05%) of lifetime volume traded.In the live census this is only a 'market-maker-like' label, not a removal. | 5 bps • | Removes | By hand | Starting value first guess |
| Old full screen: green perp month Runs by hand onlyIn the old full screen with its portfolio pass, perp profit over the last month had to be above zero.Not part of the live screen, and the next census refresh re-judges anyone this removed. | — | Removes | By hand | Starting value |
| Old full screen: green perp week Runs by hand onlyIn the old full screen with its portfolio pass, perp profit over the last week had to be above zero.Not part of the live screen, and the next census refresh re-judges anyone this removed. | — | Removes | By hand | Starting value |
| Old full screen: winning days Runs by hand onlyIn the old full screen with its portfolio pass, at least half of the days with perp profit or loss in the last month had to be winning days.Not part of the live screen, and the next census refresh re-judges anyone this removed. | 50% • | Removes | By hand | Starting value |
| Old full screen: best-day share Runs by hand onlyIn the old full screen with its portfolio pass, no single day could supply more than 40% of the month's perp profit.Not part of the live screen, and the next census refresh re-judges anyone this removed. | 40% • | Removes | By hand | Starting value |
| One-off: perp profit check RetiredIn a one-off run in September 2026, wallets with under $500,000 of perp-only profit were moved to the long-term list as spot holders instead of being judged as traders.No longer applied: the daily screen now keeps spot traders as a type of their own. A spot holder moved by these runs that passes the screen joins the deep-check queue; one that fails stays on the long-term list. | $500,000 | Routes | Once, by hand (12 Sep 2026) | Owner's decision |
| One-off: 50% return gate RetiredIn the same September 2026 one-off runs, wallets kept by the perp check also needed an all-time leaderboard return of at least 50%. Those below were removed for low return.No longer applied: the daily screen has no return rule and screens these wallets again. One that passes joins the deep-check queue; one that fails takes the screen's own reason. | 50% • | Removes | Once, by hand (12 Sep 2026) | Owner's decision |
| Off-leaderboard wallets: $1M perp account Runs by hand onlyA wallet found outside the leaderboard also needs a perp account worth at least $1,000,000.Only runs by hand, and it counts the perp account alone. | $1,000,000 • | Removes | By hand | Starting value |
| Off-leaderboard wallets: green month and week Runs by hand onlyA wallet found outside the leaderboard must also be in profit on perps over both the last month and the last week.Only runs by hand. | $0 | Removes | By hand | Starting value |
| Off-leaderboard wallets: winning days Runs by hand onlyFor a wallet found outside the leaderboard, at least half of the days with perp profit or loss in the last month must be winning days.Only runs by hand. | 50% • | Removes | By hand | Starting value |
| One-day month (wallets found off the leaderboard) Runs by hand onlyA wallet found outside the leaderboard, for example on the live trade feed, is removed if its best day made more than 40% of its perp profit over the last month.Exactly 40% passes; leaderboard wallets are not tested on this. | 40% • | Removes | By hand | Starting value |
| Profit per dollar traded (wallets found off the leaderboard) Runs by hand onlyA wallet found outside the leaderboard is removed if its all-time perp profit is less than 5 basis points (0.05%) of its all-time perp volume.Leaderboard wallets face a similar test later, in the deep check. | 5 bps • | Removes | By hand | Starting value first guess |
| Screen re-run dailyOnce a day the screen runs, most profitable first, over wallets with $500K+ all-time profit that it has not screened yet, over its own rejects (account under $1M, or no trade in 90 days) and wallets that failed it while holding a deeper verdict, once their 30-day re-check date has passed, and over any typed wallet waiting in the queue without a pass. It makes at most 900 exchange requests a run, and the rest wait for the next day.Wallets that pass are not screened again, so a passer that later goes quiet or drops under $1M keeps its pass; the top-trader lists (a trade in 30 days or a running trade) and the 'Traded within' filter still drop a quiet one. | 24 h; up to 900 calls | Schedule | Every 24 h | Starting value |
4 · Who is behind the account Vaults, sub-accounts, stakers, hedgers, market makers. · 21 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| One-off: vaults and known addresses (12 Sep 2026) RetiredA one-off run on 12 Sep 2026 removed vault accounts, vault leaders and known exchange, bridge, deployer and other protocol addresses from the queue. It used saved address lists and made no exchange requests.The deep check repeats these tests. The daily screen leaves these removals in place, whether such a wallet passes it or not. | — | Removes | Once, by hand (12 Sep 2026) | Owner's decision |
| Who the deep check takes Runs by hand onlyThe deep check takes only wallets that passed the daily screen, plus wallets with no leaderboard profit, which the screen can never judge and which only a by-hand path adds (the owner of a sub-account, an address added by hand, or a wallet admitted by the hand-run discovery check). A wallet the screen failed or has not reached yet waits for the screen.Favourites are no exception: the monitoring run already checks them in full. A passer whose latest screen failed it waits for the screen as well. A deep-check reject put back by a re-check, or a wallet left by an unfinished download, keeps its pass and is taken again. Whoever runs the check by hand can take every queued wallet with an override, or name addresses directly. | — | Schedule | By hand | Owner's decision |
| Deep-check queue order Runs by hand onlyThe deep check works through its queue in groups: perp traders first, then spot traders, then stakers, then wallets the screen has not typed yet. Within each group, wallets that traded in the last 90 days go first, then those with the largest all-time profit.A wallet with no 90-day reading is ordered by the leaderboard's 30-day activity instead; only a wallet with neither reading counts as active. | 90 days | Schedule | By hand | Owner's decision |
| Deep-check batch and request budget Runs by hand onlyEach deep-check run takes up to 20 wallets and stops once it has made 600 exchange requests, retries included. Wallets it did not reach wait for the next run.If the limit is hit part-way through one of the wallet's histories (fills, TWAP slices, ledger, funding, the account-value series), the wallet is not scored: its row records which history stopped and how many rows had arrived, it stays in the queue unchanged, and the run ends there. The next run downloads it again. | 600 | Schedule | By hand | Starting value |
| Failed checks retry Runs by hand onlyIf a wallet's deep check fails during its first account lookups or while being scored, the wallet stays in the queue and is tried again the next time the check is run.A download that gives up part-way through one of the wallet's histories, after its retries, is handled the same way: nothing is scored, the row records the unfinished download, and the wallet is tried again on the next run. | — | Schedule | By hand | Starting value |
| Quick identity checks first Runs by hand onlyBefore any trade history is downloaded, the vault, vault-leader, known-address, staker and basis-hedge checks run on four account lookups and saved address lists. A wallet that fails one of them is removed without a full rebuild.The market-maker, carry-desk, volume-farmer and liquidation checks run after the rebuild. | 4 | Schedule | By hand | Starting value |
| Sub-accounts and API wallets Runs by hand onlyIf the exchange says an address is a sub-account or an API (agent) wallet, that address is removed and the account that owns it is queued for its own deep check, unless the owner is already on file.The sub-account's own trades are not added to the owner's record. Its row is tagged with the owner, and the link is recorded as certain. Every deep-check path does this before downloading anything: the crawl, the scheduler, and a one-wallet run that writes to the database. | — | Routes | By hand | Starting value |
| Owner tagged on sub-accounts Runs by hand onlyWhen an account's deep check lists its sub-accounts, each sub-account already on file is tagged with that owner, and the link between them is recorded as certain.This happens for accounts that get past the quick identity checks and finish their download, and it never adds new wallets. Only the crawl records the sub-accounts an owner lists; the scheduler check does not. Separately, whenever the crawl or the scheduler checks a sub-account or API wallet itself, its row is tagged with the owner and the link is recorded as certain. | — | Labels | By hand | Starting value |
| Vault accounts Runs by hand onlyAn address that the exchange reports as a vault (a pool of depositors' money) is removed. | — | Removes | By hand | Starting value |
| Vault leaders Runs by hand onlyAn address that leads at least one open vault, of any size, is removed because it trades other people's money.The vault list was saved on 11 Sep 2026, so vaults opened since then are not seen. | — | Removes | By hand | Starting value |
| Known exchange and protocol addresses Runs by hand onlyAddresses publicly labelled as an exchange, bridge, vault, token deployer, treasury, airdrop, protocol or liquidator are removed. So are the operators of builder perp exchanges and Hyperliquid's own system addresses.The builder-exchange list was saved on 11 Sep 2026 and is not refreshed. | 8 label kinds • | Removes | By hand | Starting value |
| Staker Runs by hand onlyAn account is removed if staked HYPE makes up half or more of its net worth.Skipped when the HYPE price cannot be read. | 50% • | Removes | By hand | Starting value |
| What counts in net worth Runs by hand onlyFor the staker test, net worth is perp account value, plus spot stablecoins (USDC, USDT, USDH, USDE) at $1, plus spot HYPE at the live HYPE price, plus the staked HYPE itself. Every other spot token counts as zero.Leaving other tokens and vault deposits out can make a wallet look more staked than it is. | — | Measures | By hand | Starting value |
| Basis hedge Runs by hand onlyAn account is removed if a spot holding covers half or more of a short perp position in the same coin, because that is a hedged book, not a directional bet.Judged on current positions only, and only for a perp short against spot held, not the reverse. | 50% | Removes | By hand | Starting value |
| Matching spot tokens to perp coins Runs by hand onlyWrapped spot tokens are matched to their perp coin (UBTC = BTC, UETH = ETH, USOL = SOL, UXPL = XPL, UZEC = ZEC, UBNB = BNB, UDOGE = DOGE, UPUMP = PUMP, UFART = FARTCOIN, USDT0 = USDT). A builder-exchange perp such as xyz:GOOGL is matched as GOOGL.Used by the basis-hedge test and to price stablecoins. | 10 aliases • | Measures | By hand | Starting value |
| Market maker Runs by hand onlyAn account is removed if more than 80% of its perp fills were passive (maker) fills.Counted by number of fills, not size, over the history the exchange serves. | 80% • | Removes | By hand | Starting value |
| Carry desk Runs by hand onlyAn account is removed if its net funding payments come to more than half of its profit or loss over the period its trade history covers; on a re-check, over the last 30 days.A trader who is close to flat over that period can trip it with modest funding. | 50% • | Removes | By hand | Starting value |
| Volume farmer Runs by hand onlyAn account is removed if its all-time profit is less than 5 basis points (0.05%) of all the volume it has ever traded.Uses the leaderboard's all-time figures, spot included, and is skipped for wallets with no leaderboard row. | 5 bps • | Removes | By hand | Starting value first guess |
| Ever liquidated Runs by hand onlyAn account is removed if its own ledger shows it has ever been liquidated.Read from the account's ledger back to its start, not only from its recent trades. | yes • | Removes | By hand | Starting value |
| Hidden hedge (not active) Not enforcedThis check is meant to remove accounts whose returns barely move, which is a sign of a hedge held on another venue. No floor has been set, so it currently removes no one. | not set • | Measures | By hand | Starting value first guess |
| Which reason a wallet shows Runs by hand onlyWhen several checks fail, the wallet is filed under the first one in this order: sub-account or API wallet, vault, vault leader, known address, staker, basis hedge, unfinished download (held, not removed), scalper on its latest fills, history gap (held for retry), too many fills, market maker, carry desk, volume farmer, liquidated. Long-term routing and steps 6 to 10 come only after all of these.Only the first failed check is recorded against a removed wallet. | — | Shows | By hand | Starting value |
5 · Rebuilding every trade Turning fills into trades, and finding linked wallets. · 39 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Only perp fills are rebuilt Runs by hand onlyOnly perp fills are rebuilt into trades. Spot fills (names such as PURR/USDC, @107 or #12141) are left out, while builder-exchange perps such as xyz:GOOGL are kept.So a spot trader in the queue has few or no rebuilt trades and is usually removed for having too few. | — | Measures | By hand | Fixed |
| TWAP slices merged in Runs by hand onlyThe exchange serves TWAP slice fills separately from ordinary fills, so the two are merged and each fill is counted once by its trade id. | — | Measures | By hand | Fixed |
| Trim to the common window Runs by hand onlyWhen the rebuild finds a gap that comes only from the ordinary-fill and TWAP streams covering different periods, both streams are trimmed to the period they share and the trades are rebuilt from that.The TWAP stream's start counts as a cutoff only when 6,000 or more slices came back. | 6,000 • | Measures | By hand | Fixed |
| Same-millisecond orderingFills stamped in the same millisecond are put back in their true order by chaining each fill's starting position to the previous fill's ending position.The exchange returns fills from the same millisecond in arbitrary order. | — | Measures | Every 3 h and daily (position and spot jobs); by hand in the deep check | Fixed |
| Position cross-check Runs by hand onlyEvery fill's running position is checked against the exchange's own record, to within 0.000001 units. Any gap voids the wallet's numbers, and the wallet is held for a retry instead of being scored.Gaps caused only by the ordinary and TWAP fill streams covering different periods are trimmed away first. | 0.000001 • | Routes | By hand | Fixed |
| Unfinished downloads are not scored Runs by hand onlyIf one of a wallet's histories (fills, TWAP slices, ledger, funding, the account-value series) stops downloading part-way, because the request budget ran out or the exchange kept failing, the wallet is not scored. It keeps its state and its place in the queue, and its row records which history stopped and how many rows had arrived.The next run downloads the wallet again from the start. For a listed trader the rolling check is skipped as well, so an unfinished download never counts as a failing run. | — | Routes | By hand | Fixed |
| Scalper check on the latest fills Runs by hand onlyBefore a wallet's first full download, its 2,000 most recent fills are read. Each exit is matched to the size it closes, and fills are grouped into orders, so an order cut into many pieces counts once. The wallet is removed as a scalper if 30 or more exit orders were held under 5 minutes and they carry at least 30% of the winning profit of all exits that can be timed. Its full history is then not downloaded.An exit that closes size opened before those 2,000 fills cannot be timed and is left out, so for a busy swing trader few exits count and the check rarely fires. It runs on a wallet's first full download only, not on re-checks. | 30 exit orders • | Removes | By hand | Owner's decision first guess |
| Scalper check: share of the profit Runs by hand onlyThe fast exits have to carry at least this share of the winning profit of all timed exits in the 2,000 most recent fills. | 30% • | Removes | By hand | Owner's decision first guess |
| Scalper check: what counts as fast Runs by hand onlyAn exit order counts as fast if the size it closes was held, on average weighted by size, for less than this. | 5 min • | Measures | By hand | Owner's decision |
| Late start for busy accounts Runs by hand onlyIf the 2,000 most recent fills are fewer days than the history, the full download starts at the date that should hold about 100,000 fills at their pace, so it stays under the 120,000-fill limit. A wallet with fewer than 2,000 fills is downloaded from the start.Trades open at that date are left out of the statistics, like any trade open when the history begins. The track-record length still comes from the account's value history, so a late start does not make a wallet look younger, but it leaves fewer trades for the checks in steps 6 to 10. | 100,000 • | Measures | By hand | Owner's decision first guess |
| Too many fills to rebuild Runs by hand onlyAn account is removed as too deep to rebuild if its fill download still reaches about 120,000 fills, even after starting late enough for about 100,000 at its recent pace.That happens when it traded faster over the downloaded stretch than in its 2,000 most recent fills. The test counts fills, not trading decisions: an order that fills in many small pieces counts many times, so a swing trader with heavily sliced orders can be removed by it, not only a high-frequency trader or a market maker. | 120,000 | Removes | By hand | Starting value |
| Fills to orders Runs by hand onlyFills that share an order id make up one order. | — | Measures | By hand | Starting value |
| Orders to decisions Runs by hand onlyWithin a trade, orders on the same side (both adding or both reducing) that come within 2 minutes of each other, or that share an order id, count as one decision.The 2 minutes run from the decision's latest fill, so a steady stream of orders stays one decision however long it lasts. | 2 min • | Measures | By hand | Starting value first guess |
| Decisions to trades Runs by hand onlyA trade runs from flat to flat in one coin. Adding to or reducing a position stays inside the same trade, and a close never starts a new one.A trade's profit is the realised P/L the exchange reports on its fills, and fees are tracked separately. | — | Measures | By hand | Starting value |
| Flips are two trades Runs by hand onlyA fill that flips a position from long to short, or back, ends one trade and starts another at that fill.The flipping fill's realised P/L goes to the trade it closes, and its fee is split by size. | — | Measures | By hand | Starting value |
| Re-checking a wallet that has stored trades Runs by hand onlyWhen the deep check is run again on a wallet whose trades are already stored, it downloads only fills from one day before the last stored fill, puts the stored fills back in front and rebuilds the whole history.If the new download no longer reaches back to the stored fills (a busy account is served only its last two months or so) and the two do not line up, the stored copy is set aside and the wallet is rebuilt from the new download, which then holds everything the exchange still serves. Funding, and the profit it is compared with in the carry-desk test, are read over the last 30 days, or the whole stored history if it is shorter. | 1 day • | Measures | By hand | Starting value |
| Funding window on a re-check Runs by hand onlyWhen a wallet with stored trades is checked again, the carry-desk test compares funding with the account's profit over the same last 30 days, or over the whole stored history if it is shorter.A wallet's first check uses the whole period its fill history covers. Before this, a re-check compared only the few days since the last check, where two flat days with funding paid could remove a trader as a carry desk. | 30 days • | Measures | By hand | Owner's decision |
| Trades open before history starts Runs by hand onlyA position already open when the served history begins is marked incomplete and left out of the trade count, the profit statistics and hold times. Later trades in the same coin count normally.It still counts toward history length, the liquidation check and long-term routing. | — | Measures | By hand | Fixed |
| Open trades not counted Runs by hand onlyA trade still open at the end of the served history is not counted as a closed trade, and it is left out of the trade count and the profit statistics.It still counts toward history length, the liquidation check and long-term routing. | — | Measures | By hand | Starting value |
| How hold time is measured Runs by hand onlyHold time matches each unit entered to the exit that later closed it (first in, first out) and takes the median weighted by dollar size, so a small leftover tail barely moves it.Units opened before the served history begins never count, and the step 7 style checks also use the plain first-entry-to-flat hold. | — | Measures | By hand | Starting value |
| Averaging down (reported only) Runs by hand onlyAveraging down means adding to a position at a worse price than the running average entry (lower for a long, higher for a short). It is reported per trade and never used to remove anyone.The first entry of a trade never counts as an add-on. | — | Measures | By hand | Starting value |
| Execution style label Not enforcedEach decision is labelled by how it was executed. TWAP if any fill came from a TWAP order; single for one fill; fragmented for one order split into many fills; sliced for several orders whose fills are near-equal in size (variation under 15%) and evenly spaced (variation under 50%); otherwise scale-in.Not shown anywhere and not used by any check. | 0.15 / 0.5 | Labels | By hand | Starting value |
| Whose liquidation it was Runs by hand onlyA liquidation fill counts against a wallet only if that wallet was the one liquidated, not when it took the other side of someone else's liquidation.A liquidation fill that does not say who was liquidated is counted as the wallet's own, to be safe. | — | Measures | By hand | Fixed |
| Known addresses cut from the link graph Runs by hand onlyAddresses on the known-address list are cut out of the link graph first, so they cannot join unrelated wallets into one group. The list covers exchanges, bridges, vaults, deployers, treasuries, airdrops, protocols, liquidators, builder-exchange operators and Hyperliquid system addresses. | 8 label kinds • | Measures | By hand | Starting value |
| Busy addresses cut from the link graph Runs by hand onlyAny address with more than 50 transfer or sub-account partners is treated as a hub, like an exchange or bridge, and left out when wallets are grouped.Partners are counted only among the transfers recorded from wallets we have fully reviewed. | 50 • | Measures | By hand | Starting value first guess |
| Transfers link wallets Runs by hand onlyTwo wallets that sent funds directly to each other, even once, are treated as one owner and grouped together.Only transfers found in the ledgers of fully reviewed wallets count, and hub addresses are ignored. | 1 | Labels | By hand | Starting value |
| Sub-accounts grouped with owner Runs by hand onlyA sub-account, or an agent wallet that trades on another account's behalf, is grouped with the account that owns it.Unless the owner is itself treated as a hub, for example one with more than 50 linked accounts. | — | Labels | By hand | Starting value |
| Weaker signals need two kinds Runs by hand onlyWeaker signals group two wallets only when at least two different kinds agree. Today the only kinds are trading against each other and trading in lockstep, so both are needed, and a pair grouped this way is always also caught by the wash-trading rule.Shared-referrer and shared-builder signals are not collected yet. | 2 | Removes | By hand | Starting value |
| Trading against each other: count Runs by hand onlyTwo wallets must have been on opposite sides of at least 20 trades before they can be linked as trading against each other. Those trades must also be a large enough share of their activity.Counted on the live trade feed of the ~40 busiest coins, which records only when someone starts it by hand, and only for trades where both wallets had already been admitted to the feed ($25,000 in one fill or $500,000 in 7 days). | 20 • | Measures | By hand | Starting value first guess |
| Trading against each other: share Runs by hand onlyThe trades two wallets made against each other must be at least 10% of either wallet's trades on the live feed.Pair counts are kept for all time but each wallet's total covers only the last two weeks, so this test loosens over time. | 10% • | Measures | By hand | Starting value first guess |
| Copy echo window Runs by hand onlyAn echo is a trade that takes liquidity in the same coin and direction within 2 seconds after another wallet's liquidity-taking trade.Resting orders are ignored, so two wallets filled by the same large order do not look like copying. | 2 s • | Measures | By hand | Starting value |
| Copy echo: minimum count Runs by hand onlyA copy link needs at least 20 echoes. | 20 • | Measures | By hand | Starting value first guess |
| Copy echo: minimum share Runs by hand onlyAt least 30% of the follower's liquidity-taking trades must be echoes of the other wallet. | 30% • | Measures | By hand | Starting value first guess |
| Copy echo: beyond chance Runs by hand onlyEchoes must be at least 5 times more common than chance, given how often the other wallet trades that coin in that direction.Chance is estimated by spreading the other wallet's recorded trades over the whole span of data on file, not over the time that wallet was actually watched, so for a wallet seen only on the 14-day trade feed this test is much weaker than 5 times chance. | 5 • | Measures | By hand | Starting value first guess |
| Wash trading Runs by hand onlyIf two wallets in the same group have traded against each other often enough to be linked (at least 20 trades, and at least 10% of either one's trades), every wallet in the group is removed as wash trading.Never re-checked. | — | Removes | By hand (after a manual review step) | Starting value |
| Copy-follower Runs by hand onlyA wallet that keeps echoing another wallet's trades, with no transfer or sub-account link between them and no echo the other way, is removed as a copy-follower. The wallet it copies is added for screening if it is new.Never re-checked. | — | Removes | By hand (after a manual review step) | Starting value |
| Groups do not merge scores Runs by hand onlyWallets in the same group share a group tag, but each one is still judged on its own trades. Group results are not combined. | — | Labels | By hand (after a manual review step) | Starting value |
| When grouping runs Runs by hand onlyGrouping and the wash-trading and copy-follower checks are not on a schedule. They run only when someone starts them by hand, and they change a wallet's status only when that run is set to save its results.The live trade feed they rely on is also started by hand. | — | Schedule | By hand | Starting value |
| Two-way echo is not copying Runs by hand onlyIf two wallets echo each other's trades in both directions, they are treated as two traders reacting to the same signal, not as a copier and a leader, and neither is removed as a copy-follower.If they also trade against each other often enough to be linked, they are grouped, and because that link sits inside the group, the whole group is removed as wash trading. | — | Measures | By hand (after a manual review step) | Starting value |
6 · One lucky bet? Is the profit spread over many independent trades? · 5 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Enough closed trades Runs by hand onlyA wallet needs at least 50 closed trades, each opened from flat and closed back to flat, in the trade history the exchange provides.Busy accounts only have about their last 50–80 days of trades available, so only those count. | 50 • | Removes | By hand | Starting value |
| Top 3 trades cap Runs by hand onlyThe three most profitable trades may make up at most 40% of the total profit from all winning trades.Losing trades are not subtracted from that total. | 40% • | Removes | By hand | Starting value |
| Effective number of winners Runs by hand onlyProfit must be spread across winning trades as evenly as if it came from at least 15 equal-sized winners.Only winning trades count, and a few very large winners pull this number down. | 15 • | Removes | By hand | Starting value |
| Profit without best trade Runs by hand onlyNet profit from closed trades must still be at least $500K after removing the single best trade.Only trades opened and closed inside the available history count, which makes this hard to pass for busy accounts. | $500,000 • | Removes | By hand | Starting value |
| Profit without best month Runs by hand onlyNet profit from closed trades must still be at least $500K after removing its best 30-day period.Periods are fixed 30-day blocks set by the date each trade closed, not calendar months. | $500,000 • | Removes | By hand | Starting value |
7 · Scalper? Does it hold long enough to be followed? · 12 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Algo signal: no sleep Runs by hand onlyAlgo signal: on a typical day with trades, the longest break between the wallet's fills is under 2 hours.Days run midnight to midnight UTC. This is one of three signals behind the Algo label. | 2 h | Labels | By hand | Starting value |
| Algo signal: round-the-clock Runs by hand onlyAlgo signal: the wallet's fills are spread almost evenly across all 24 hours of the day (a spread score above 0.95, where 1 is perfectly even).One of three signals behind the Algo label. | 0.95 | Labels | By hand | Starting value |
| Algo signal: repeated sizes Runs by hand onlyAlgo signal: more than 60% of the wallet's fills are one of its three most common dollar sizes, rounded to the nearest $100.One of three signals behind the Algo label. | 60% | Labels | By hand | Starting value |
| Algo label Runs by hand onlyA wallet is labelled Algo when at least two algo signals fire. The label is shown on its profile and never removes anyone.A planned fourth signal, reaction speed, is not measured yet, so only three can fire. | 2 | Labels | By hand | Starting value |
| Sub-minute trades cap Runs by hand onlyAt most 10% of closed trades may be held for under one minute.Hold time runs from the first entry until the position is flat again. | 10% • | Removes | By hand | Starting value |
| Scalper hold floor Runs by hand onlyThe median closed trade must be held for at least 5 minutes.Any wallet below this also fails the one-hour median rule, so on its own it changes nothing. | 5 min • | Removes | By hand | Starting value |
| Median hold at least 1 hour Runs by hand onlyThe median closed trade must be held for at least one hour.The same limit is applied to the last 60 days when a listed trader is re-checked. | 1 h • | Removes | By hand | Starting value |
| Size-weighted hold at least 1 hour Runs by hand onlyMeasuring each trade by how long each dollar of the position was held, the median closed trade must still last at least one hour.This stops a small leftover held for days from making a fast trade look slow. | 1 h • | Removes | By hand | Starting value |
| Sub-5-minute trades cap Runs by hand onlyAt most 30% of closed trades may be held for under five minutes.The same limit is applied to the last 60 days when a listed trader is re-checked. | 30% • | Removes | By hand | Starting value |
| Long-term route: few trades Runs by hand onlyA wallet with fewer than 30 closed trades and a median hold of a week or more goes to the long-term list instead of being judged as a trader.Long-term wallets skip the trading rules and are shown for their positioning, never as someone to follow. | 30 • | Routes | By hand | Starting value |
| Long-term route: week-long holds Runs by hand onlyTo go to the long-term list, a wallet's median closed trade must be held for at least 7 days, and the wallet must also have fewer than 30 closed trades.A wallet with no closed trades at all is handled by the held-through rule instead. | 7 days • | Routes | By hand | Starting value |
| Held through the whole window Runs by hand onlyA wallet with no closed flat-to-flat trade across at least 30 days of trade history goes to the long-term list. Every position must be either still open or already open when the history begins.The 30 days run from its first to its last position opened or closed, not to today. | 30 days | Routes | By hand | Starting value |
8 · Risk Drawdown, single-trade losses, liquidations. · 3 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Maximum drawdown Runs by hand onlyThe account's largest fall from a peak may be at most 35%.Deposits and withdrawals are taken out, and the exchange's history is only about weekly for older accounts, so short dips can be missed. | 35% • | Removes | By hand | Starting value |
| Worst single-trade loss Runs by hand onlyNo single closed trade may lose more than 15% of the account's value when it was opened.The account value comes from the exchange's nearest earlier snapshot, which can be several days old. | 15% • | Removes | By hand | Starting value |
| Never liquidated (fills) Runs by hand onlyA wallet is removed if any of its fills shows that it was the one liquidated.Only the trade history the exchange provides is checked here, which for busy accounts is about the last 50–80 days. | — | Removes | By hand | Starting value |
9 · Track record How long a record has to be before it is judged. · 3 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| How long the record is Runs by hand onlyA wallet's track record is the longer of two spans: the exchange's account-value history, or the span of its recorded trades.So an older account can count as established even when only its last couple of months of trades are available. | — | Measures | By hand | Starting value |
| Too young Runs by hand onlyA wallet with less than 90 days of record is removed as too young. It is due for a re-check on the day it turns 90 and is not kept on a watchlist.If it also failed an earlier check, that check's re-check date applies instead. | 90 days • | Removes | By hand | Starting value |
| Candidate or proven Runs by hand onlyA wallet that passes with 90 to 270 days of record is a candidate: a watchlist, never something to size. With 270 days or more it is proven, and only proven wallets face the consistency tests. | 270 days • | Routes | By hand | Starting value |
10 · Consistency Is the edge steady, or one good season? · 25 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Profitable months Runs by hand onlyAt least 60% of its months must be profitable.A month is a fixed 30-day block with at least one closed trade, or a calendar month of the exchange's profit figure when older trades are missing. | 60% • | Removes | By hand | Starting value |
| No single dominant month Runs by hand onlyIts best month may be at most 40% of its total profit.Uses the same months as the profitable-months test and is not applied when the total is a loss. | 40% • | Removes | By hand | Starting value |
| Months when older trades are missing Runs by hand onlyWhen older trades are missing, the monthly tests use calendar months of the exchange's own profit figure, including open positions, instead of closed trades. Trades count as missing when a position was already open where the available trade history begins, or when that history is more than 7 days shorter than the account history. | 7 days | Measures | By hand | Starting value |
| Rolling 30-day return positive Runs by hand onlyIts return over the trailing 30 days must be positive at least 55% of the time.Checked at each account-value reading. When deposits and withdrawals make returns unreliable, the 30-day change in profit is used instead. | 55% • | Removes | By hand | Starting value |
| Edge not fading Runs by hand onlyThe second half of its record must be at least half as good as the first half, measured by risk-adjusted return (Sharpe ratio).Only applied when the record has at least 40 account-value readings. | 0.5 • | Removes | By hand | Starting value |
| Fading test needs an early edge Runs by hand onlyThe edge-fading test is skipped unless the first half of the record has a Sharpe ratio of at least 0.25. Without an early edge there is nothing to fade. | 0.25 | Measures | By hand | Starting value |
| Worst loss vs typical loss Runs by hand onlyIts worst losing trade may be at most 10 times its median losing trade. | 10 • | Removes | By hand | Starting value |
| No sizing up after wins or losses Runs by hand onlyIts trade size, as a share of the account, must not follow the previous trade's result. The correlation may be at most 0.25 either way.Size is the trade's largest position divided by the account value when it opened. | 0.25 • | Removes | By hand | Starting value |
| Two-sided or one-directional Runs by hand onlyIf its less-used side has at least a quarter as many closed trades as its main side, a trader is two-sided and must pass the weaker-side test. Otherwise it is one-directional and gets the market-trend test instead. | 0.25 | Routes | By hand | Starting value |
| Enough weaker-side trades Runs by hand onlyA two-sided trader needs at least 20 closed trades on the side it trades less. Fewer is a failure. | 20 • | Removes | By hand | Starting value |
| Weaker side makes money Runs by hand onlyA two-sided trader's average closed trade on its less-used side must make money (more than $0).Checked once it has the 20 weaker-side trades. | $0 • | Removes | By hand | Starting value |
| Market-against-it test Runs by hand onlyA one-directional trader must, on average, make money on its main-side trades opened while Bitcoin's 30-day trend ran against that side. The test is decided once it has at least 10 such trades.With fewer than 10 such trades, or without Bitcoin price data, the test is not decided and removes no one. | 10 • | Removes | By hand | Starting value first guess |
| A repeated setup: trade count Runs by hand onlyIt must have at least one repeated setup with 20 or more closed trades that make money on average and pass the significance test.A setup means the same coin group, side, hold length, market volatility and size band. | 20 • | Removes | By hand | Starting value |
| A repeated setup: significance Runs by hand onlyThat setup's average profit must have a t-statistic of at least 2, so it is unlikely to be luck.The t-statistic is the average profit divided by its standard error. | 2.0 • | Removes | By hand | Starting value |
| Setup hold lengths Runs by hand onlyFor setups, hold time is grouped as under 1 hour, 1 to 4 hours, 4 to 24 hours, 1 to 7 days, and over 7 days.Hold time is the trade's size-weighted median hold. | <1h, 1-4h, 4-24h, 1-7d, >7d • | Measures | By hand | Starting value |
| Setup size bands Runs by hand onlyFor setups, trade size is grouped as under 5%, 5 to 20%, 20 to 50%, and over 50% of the account value when the trade opened.Trades with no known account value form their own group. | <5%, 5-20%, 20-50%, >50% • | Measures | By hand | Starting value |
| Setup coin groups Runs by hand onlyFor setups, coins are grouped as majors (BTC, ETH, SOL), large alts (the rest of the top 40 by daily volume), small alts, and builder-deployed markets.The top 40 comes from a volume snapshot of 11 September 2026 that is not refreshed. | BTC, ETH, SOL • | Measures | By hand | Starting value |
| Sharpe confidence range Runs by hand onlyA 95% confidence range for the Sharpe ratio of its trade results, scaled to a year by how many trades it makes, is reported by resampling trades in blocks of 10 in a row. It never removes anyone.Reported for proven wallets only. | 10 • | Measures | By hand | Starting value first guess |
| Sharpe confidence range: resamples Runs by hand onlyThe confidence range is built from 1,000 resamples. | 1,000 • | Measures | By hand | Starting value |
| Beta to Bitcoin Runs by hand onlyHow much the account's returns move with Bitcoin (its beta) is reported as a cross-check once there are at least 10 comparable periods. It never removes anyone. | 10 | Measures | By hand | Starting value |
| Market trend Runs by hand onlyBitcoin's trend on a day is up if its close is at or above its close 30 days earlier, and down otherwise.The Bitcoin prices on file run from July 2024 to 11 September 2026 and are not refreshed. Trades before August 2024 get no trend, and later trades take the last day's. | 30 days | Measures | By hand | Starting value |
| Market volatility bands Runs by hand onlyBitcoin's trailing 30-day volatility is split into low, middle and high thirds, with cut points set from the stored Bitcoin price history.Used to group setups, and as part of the market regimes the long-term list counts before it scores a holder's timing. | — | Measures | By hand | Starting value |
| Market data required Runs by hand onlyThe market-trend test, the repeated-setup test, the Sharpe confidence range and the Bitcoin beta run only when at least 60 days of Bitcoin daily prices are on file.Without that data, one-directional traders are flagged as untested rather than removed. | 60 days | Measures | By hand | Starting value |
| Sizing test needs 10 trade pairs Runs by hand onlyThe test for sizing up after wins or losses runs only when at least 10 pairs of consecutive trades have a known account value. With fewer pairs it is skipped and does not count against the trader. | 10 | Measures | By hand | Starting value |
| Sharpe range needs 20 trades Runs by hand onlyThe Sharpe confidence range is computed only with at least 20 closed trades. Below that it is left blank. | 20 | Measures | By hand | Starting value first guess |
11 · Forward test Does the list predict anything? · 15 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| The frozen list Runs by hand onlyOn freeze day every candidate and proven trader is frozen into the test and ranked by t-statistic, with the Sharpe ratio breaking ties. Rank 1 is best.Ranks are fixed on freeze day, and traders marked as slipping at that time are left out. | — | Measures | By hand | Starting value |
| Favourites group Runs by hand onlyFavourites are frozen as their own group, whatever their status and without a rank. They never count toward the list's score.A favourite that is also on the list counts only as a favourite. | — | Measures | By hand | Starting value |
| Control group Runs by hand onlyUp to 50 random wallets are frozen next to the list as a control group. Each was fully rebuilt and then removed by the lucky-bet, scalper, risk or consistency checks.Wallets removed only for being too young, or for slipping after being listed, are not in the pool. | 50 | Measures | By hand | Starting value |
| Forward window Not enforcedEach frozen group is meant to be followed for 90 days.Not enforced yet: tracking has no end date, so a report covers however many days have passed. | 90 days | Measures | By hand | Starting value |
| How forward return is measured Runs by hand onlyForward return is the time-weighted return when no period lost everything and deposits and withdrawals are typically at most 50% of the account per period. Otherwise it is the profit since freeze divided by the account value at freeze.For accounts with heavy cash flows the forward return may be blank. | 0.50 • | Measures | By hand | Starting value first guess |
| Still active Runs by hand onlyA frozen trader counts as still active if it traded perps in the last 7 days. | 7 days | Measures | By hand | Starting value |
| Blow-up count Runs by hand onlyThe report counts a frozen trader as blown up when, since freeze, its profit fell from its peak by more than 50% of the account it was running.This is only a tally: blown-up traders stay in the test. | 50% | Shows | By hand | Starting value |
| Copy delay Runs by hand onlyThe simulated copier repeats each of the trader's trades 2 seconds later at the same size. Its result is measured against the trader's account value at freeze.Runs only for listed traders and favourites whose trades since freeze are stored, and copied positions still open are not counted. | 2,000 ms • | Measures | By hand | Starting value |
| Copier pays taker fees Runs by hand onlyEvery simulated copy trade pays Hyperliquid's base taker fee of 4.5 bps (0.045%). | 4.5 bps • | Measures | By hand | Fixed |
| Copy price from real trades Runs by hand onlyThe copier's price is the first recorded trade in that coin at least 2 seconds after the trader's fill, if it came within 60 seconds of the trader's fill.Recorded trades come from the trade feed, which runs only when started by hand and keeps trades for 14 days; when it is not running, or the trade is older, the fallback price is used. | 60 s | Measures | By hand | Starting value |
| Fallback copy price Runs by hand onlyWhen no recorded trade price is available, the copier gets the trader's own price moved 2 bps (0.02%) against it. | 2 bps • | Measures | By hand | Starting value first guess |
| Minimum sample for the verdict Runs by hand onlyThe rank test checks whether a better rank at freeze meant a better forward return. It is computed only once at least 5 listed traders have a forward return. | 5 | Measures | By hand | Starting value |
| Deploy or do not deploy Runs by hand onlyThe forward test's verdict is 'do not deploy' unless the rank correlation between selection rank and later return is at least 0.10 in size and its 95% range excludes zero.A clearly negative correlation also avoids 'do not deploy', even though it means the ranking works backwards. | 0.10 | Shows | By hand | Starting value |
| Is the list one bet? Runs by hand onlyIf the listed traders' open positions look alike on average (average similarity above 0.5, by coin, direction and size relative to the account), the list is flagged as one bet rather than many independent traders.Only Candidate and Proven traders with open positions are compared; nobody is removed. | 0.50 • | Measures | By hand | Starting value first guess |
| Does the list fit the market? Runs by hand onlyA listed trader's position is flagged when it is more than 5% of that coin's open interest, and the list's combined share of open interest is reported for its biggest coins.Nobody is removed. | 5% | Measures | By hand | Starting value first guess |
Staying on the list How a listed trader is re-checked. · 15 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Who is monitored Runs by hand onlyEvery listed trader (Candidate, Proven or Slipping) and every favourite is pulled again and fully re-checked, least recently updated first, up to 200 wallets a run.A run also stops once it has used 800 exchange requests. | 200 | Schedule | By hand | Starting value |
| Topping up the trade history Runs by hand onlyEach re-check reads the trader's fills again from one day before the newest fill already stored, and adds them to the stored history. | 1 day | Measures | By hand | Starting value |
| Rolling 60-day window Runs by hand onlyThe monitoring check looks only at trades closed in the last 60 days.Trades still open, or whose start is missing from the exchange's history, are left out; the drawdown inside the window is measured but not used. | 60 days • | Measures | By hand | Starting value |
| Still trading Runs by hand onlyA listed trader that closed no trade in the last 60 days fails the check as dormant.This includes a trader holding one position open the whole time; it is a strike, not an immediate removal. | 60 days • | Removes | By hand | Starting value |
| Still profitable Runs by hand onlyThe trades closed in the last 60 days must add up to a profit, counted before fees and funding.A failure counts as a strike. | $0 | Removes | By hand | Starting value |
| No drift into scalping: hold time Runs by hand onlyThe median trade closed in the last 60 days must have been held at least 1 hour, from first entry to fully closed.A failure counts as a strike. | 1 h • | Removes | By hand | Starting value |
| No drift into scalping: very short trades Runs by hand onlyNo more than 30% of the trades closed in the last 60 days may have lasted under 5 minutes.A failure counts as a strike. | 30% • | Removes | By hand | Starting value |
| No liquidation Runs by hand onlyIf the trader was liquidated in any trade closed in the last 60 days, the check fails.In practice the full re-check run alongside removes a liquidated trader at once. | — | Removes | By hand | Starting value |
| Change of style Runs by hand onlyIf the median hold of trades closed in the last 60 days is more than 4 times longer or shorter than before the window, the check fails as a change of style.Tested only with at least 20 earlier closed trades and 10 in the window; it counts as a strike and does not restart the track record. | 4x • | Removes | By hand | Starting value first guess |
| Two strikes and out Runs by hand onlyTwo failed 60-day checks in a row remove the trader, with the reason 'Dropped out'.A passing check in between resets the count. | 2 • | Removes | By hand | Starting value |
| First strike: Slipping Runs by hand onlyOne failed 60-day check moves a Candidate or Proven trader to Slipping, where it stays listed and monitored.If the full re-check rejects it in the same run, it is removed instead. | 1 | Labels | By hand | Starting value |
| Full re-check on every run Runs by hand onlyEach monitoring run also repeats the full evaluation. An identity finding (vault, staker, market maker, exchange address, past liquidation and the like) or a history too deep to rebuild removes the trader at once under that reason; a failed lucky-bet, scalper, risk or consistency check removes it too, but when that run is also its second failed 60-day check in a row the reason recorded is 'Dropped out'.The quick identity checks run first, before anything is downloaded: a listed trader that turns out to be a sub-account or API wallet is removed and its owner is queued unless it is already on file, and a vault, vault leader, known address, staker or basis hedge is removed under that reason. The wash-trading and copy-follower checks are not repeated here. | — | Removes | By hand | Starting value |
| Favourites get no exemption Runs by hand onlyA favourite is always re-checked by monitoring, whatever its state, and skips no check. The 60-day strike rule applies to it only while it is Candidate, Proven or Slipping.A removed favourite is simply re-judged each run and can return to the list, even after a wash-trading or copy-follower finding, which is not repeated. | — | Schedule | By hand | Starting value |
| Planned job cadence Runs by hand onlyWhen the job loop is running, re-checks and monitoring run about once a day and discovery every 2 days.The loop is not running on the live site; new wallets arrive through the 48-hour census refresh and the daily screen instead. | 1 day / 2 days | Schedule | By hand | Starting value |
| A clean window clears the strike Runs by hand onlyIf a Slipping trader's next 60-day check passes, its strike count goes back to zero and it takes whatever standing the full re-check gives.If the full re-check is inconclusive because of a gap in the trade history, it returns as Candidate, even if it was Proven before. | — | Routes | By hand | Starting value |
When a removed wallet is looked at again The wait before each reason is re-tested. · 36 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Screen removals keep their verdictA wallet the screen removed for an account under $1M or for no trade in 90 days keeps that verdict and its re-check date through census refreshes; only its leaderboard numbers are updated, even if its profit later drops under $500K.It is screened again only once that date has passed and its lifetime profit is $500K or more. | — | Schedule | Every 48 h | Fixed |
| Under the profit floorA wallet under the $500K lifetime-profit floor never gets a re-check date; instead every census refresh judges it again, and once it is above the floor it re-enters the pipeline.If it was screened before, its earlier screen result is reused rather than run again. | never • | Schedule | Every 48 h | Starting value |
| Account under $1MA wallet removed because its account was under $1M is screened again after 30 days. Its account value and its perp/spot profit split are read again from its portfolio, and its staking rewards again; its activity counts from its last trade on record, and its fills are read only if that trade is over 90 days old or there is none.Only while its lifetime profit is still at least $500K. | 30 days • | Schedule | Daily | Starting value |
| No trade in 90 daysA wallet removed for no trade in 90 days is checked again after 30 days; only the activity check is repeated and its trader type is kept. A trade on record in the last 90 days passes it with no request; otherwise its recent fills are read.Only while its lifetime profit is still at least $500K. | 30 days • | Schedule | Daily | Starting value |
| Sub-account or API wallet Runs by hand onlyA sub-account or API wallet set aside in favour of the account that owns it is due for another look after 30 days. | 30 | Schedule | By hand | Starting value |
| Identity removals Runs by hand onlyA wallet removed as a vault, a vault leader, an exchange or protocol address, a staker, a basis hedge, a market maker, a carry desk, a volume farmer or hedged elsewhere is due for another look after 180 days.The hedged-elsewhere check has no threshold set yet, so it never removes anyone. | 180 • | Schedule | By hand | Starting value |
| Which failure names the removal Runs by hand onlyWhen several checks fail, the wallet is labelled with the first failure in this order: one lucky bet, scalper, risk, track record, consistency. An identity finding, the scalper check on the latest fills, or a history too deep to rebuild outranks all of them.Only that first reason is kept, and it sets when the wallet is due for another look. | — | Labels | By hand | Starting value |
| Re-check queue Runs by hand onlyWhen a removed wallet's re-check date passes, it goes back for a fresh evaluation by the path that removed it: a deep-check removal back into the deep-check queue, a removal by the hand-run discovery check back to discovery.The daily screen's own removals (an account under $1M, no trade in 90 days) are re-checked automatically by the screen, and this job leaves them alone; every other date waits for this job to be run by hand. | — | Schedule | By hand | Starting value |
| Default re-check Runs by hand onlyA removal whose reason has no cadence of its own is due for another look after 30 days.This covers, for example, sub-accounts and API wallets. | 30 days | Schedule | By hand | Starting value |
| Too young: re-check at 90 days Runs by hand onlyA trader removed only because it has under 90 days of history is due for another look once it reaches 90 days. | 90 days | Schedule | By hand | Starting value |
| Trades did not add up Runs by hand onlyIf the rebuilt positions do not match the exchange's, the wallet is not judged: it stays in the evaluation queue and is due to be retried after 3 days.Retries happen only when someone runs the crawl, which takes queued wallets whatever the date. | 3 days • | Schedule | By hand | Starting value |
| One lucky bet (step 6) Runs by hand onlyA wallet removed at the one-lucky-bet step is due for another look after 30 days. | 30 days • | Schedule | By hand | Starting value |
| Scalper (step 7)A wallet removed as a scalper is never re-checked.This includes a wallet removed by the scalper check on its latest fills (step 5). Favourites are the exception: monitoring re-judges them whatever their state. | never • | Schedule | Never | Starting value |
| Risk (step 8) Runs by hand onlyA wallet removed at the risk step (deep drawdown, one outsized loss or a liquidation) is due for another look after 60 days. | 60 days • | Schedule | By hand | Starting value |
| Consistency (step 10) Runs by hand onlyA wallet removed at the consistency step is due for another look after 30 days. | 30 days • | Schedule | By hand | Starting value |
| Dropped out after two strikes Runs by hand onlyA shortlisted wallet that fails the rolling 60-day check on two monitoring runs in a row is removed. It comes up for a re-check after 30 days.The monitoring runs that count the strikes are also started by hand. | 30 days • | Schedule | By hand | Starting value |
| Too many fills to rebuild: when it is looked at again Runs by hand onlyA wallet rejected because it has more fills than the checker can page through comes up for a re-check after 60 days.The download first starts late enough for about 100,000 fills at the account's recent pace, so this is left for accounts that traded faster over that stretch. It counts fills, not decisions. | 60 days • | Schedule | By hand | Starting value |
| Liquidated Runs by hand onlyA wallet rejected because its account history shows a liquidation comes up for a re-check after 60 days.The re-check reads the whole history again, so a past liquidation gets it rejected again. | 60 days • | Schedule | By hand | Starting value |
| Staker: when it is looked at again Runs by hand onlyA wallet rejected as a HYPE staker comes up for a re-check after 180 days. | 180 days • | Schedule | By hand | Starting value |
| Vault account Runs by hand onlyAn address rejected because it is itself a vault comes up for a re-check after 180 days. | 180 days • | Schedule | By hand | Starting value |
| Vault leader Runs by hand onlyA wallet rejected because it runs a vault comes up for a re-check after 180 days. | 180 days • | Schedule | By hand | Starting value |
| Exchange, bridge or deployer Runs by hand onlyAn address rejected as a known exchange, bridge, token deployer or similar service comes up for a re-check after 180 days. | 180 days • | Schedule | By hand | Starting value |
| Market maker: when it is looked at again Runs by hand onlyA wallet rejected as a market maker comes up for a re-check after 180 days. | 180 days • | Schedule | By hand | Starting value |
| Hedged (basis) account Runs by hand onlyA wallet rejected as hedged, with spot holdings offsetting a perp short, comes up for a re-check after 180 days. | 180 days • | Schedule | By hand | Starting value |
| Funding carry Runs by hand onlyA wallet rejected because most of its profit came from funding payments comes up for a re-check after 180 days. | 180 days • | Schedule | By hand | Starting value |
| Volume farmer: when it is looked at again Runs by hand onlyA wallet rejected for making too little profit per dollar traded comes up for a re-check after 180 days. | 180 days • | Schedule | By hand | Starting value |
| Wash trading: when it is looked at again Runs by hand onlyA wallet rejected for wash trading is never re-checked.These rejections are only made when the wallet-linking check is run by hand. | never • | Schedule | Never | Starting value |
| Copy follower Runs by hand onlyA wallet rejected for copying another wallet's trades is never re-checked.These rejections are only made when the wallet-linking check is run by hand. | never • | Schedule | Never | Starting value |
| Low all-time return (leaderboard screen) Runs by hand onlyIf the older multi-rule leaderboard screen is run by hand and rejects a wallet for an all-time return under 50%, the wallet comes up for a re-check after 30 days.The automatic leaderboard refresh every 48 hours re-judges it sooner, under the current $500K-profit rule. | 30 days • | Schedule | By hand | Starting value |
| Account under $1M (leaderboard screen or discovery) Runs by hand onlyA wallet rejected for an account under $1M by the older multi-rule leaderboard screen or by new-wallet discovery comes up for a re-check after 30 days.Both of those runs are started by hand, and wallets still on the leaderboard are re-judged sooner by the automatic 48-hour refresh. | 30 days • | Schedule | By hand | Starting value |
| No profit last month (leaderboard screen) Runs by hand onlyIf the older multi-rule leaderboard screen is run by hand and rejects a wallet for making no profit over its last 30 days, the wallet comes up for a re-check after 14 days.The automatic leaderboard refresh every 48 hours re-judges it sooner, under the current $500K-profit rule. | 14 days • | Schedule | By hand | Starting value |
| No profit last week (leaderboard screen) Runs by hand onlyIf the older multi-rule leaderboard screen is run by hand and rejects a wallet for making no profit over its last 7 days, the wallet comes up for a re-check after 14 days.The automatic leaderboard refresh every 48 hours re-judges it sooner, under the current $500K-profit rule. | 14 days • | Schedule | By hand | Starting value |
| Too few winning days Runs by hand onlyA wallet rejected because fewer than half of the days with any perp profit or loss in its last month were winning days comes up for a re-check after 14 days.Only runs started by hand make this rejection, and wallets still on the leaderboard are re-judged sooner by the automatic 48-hour refresh. | 14 days • | Schedule | By hand | Starting value |
| One day made the month Runs by hand onlyA wallet rejected because its single best day made more than 40% of its last month's perp profit comes up for a re-check after 60 days.Only runs started by hand make this rejection, and wallets still on the leaderboard are re-judged sooner by the automatic 48-hour refresh. | 60 days • | Schedule | By hand | Starting value |
| Too little profit per dollar traded (leaderboard screen) Runs by hand onlyA wallet rejected by the leaderboard screen or new-wallet discovery for less than 5 cents of profit per $100 traded (0.05%) comes up for a re-check after 180 days.Only runs started by hand make this rejection, and wallets still on the leaderboard are re-judged sooner by the automatic 48-hour refresh. | 180 days • | Schedule | By hand | Starting value |
| No perp profit last month or week, or no perp data Runs by hand onlyA wallet rejected for making no perp profit over its last month or last week, or because its perp history could not be read, comes up for a re-check after 30 days.Only runs started by hand make this rejection, and wallets still on the leaderboard are re-judged sooner by the automatic 48-hour refresh. | 30 days | Schedule | By hand | Starting value |
The long-term list Holders scored on timing, never followed. · 12 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Spot profit goes to the long-term tab Runs by hand onlyA wallet with $500K+ of all-time profit, but less than $500K of it from perps (the rest came from spot or HYPE), goes to the long-term tab as a spot holder instead of being rejected.It gets no timing score until it is scored separately. | $500,000 | Routes | By hand | Owner's decision |
| Spot holder or long-duration perp holder Runs by hand onlyWhen a long-term wallet is scored, it is called a spot holder if its stablecoin balance is worth more than its open perp positions, and a long-duration perp holder otherwise. Other tokens are not counted, so a wallet holding only tokens such as HYPE is called a perp holder.The tracker groups these wallets by where their profit came from instead, so the two labels can disagree. | — | Labels | By hand | Starting value |
| What counts as an exposure change Runs by hand onlyAn exposure change is a day on which the wallet's perp position in a coin moved, net, by at least 10% of its size at the start of that day, or by any amount if the day started flat.Only perp positions count; spot buying and selling is left out. | 10% • | Measures | By hand | Starting value first guess |
| 30-day look-ahead Runs by hand onlyEach exposure change is scored against the coin's price change over the following 30 days.Changes less than 30 days old, or more than about 800 days old, are not scored. | 30 days | Measures | By hand | Starting value |
| Timing score Runs by hand onlyThe timing score is the correlation between each exposure change (how much the wallet added or cut, relative to its position) and the coin's price move over the next 30 days.A positive score means it tended to buy before prices rose and sell before they fell. | — | Measures | By hand | Starting value |
| Minimum events Runs by hand onlyA timing score is only shown when it rests on at least 8 scored exposure changes, meaning ones with a full 30 days of prices after them. | 8 • | Measures | By hand | Starting value first guess |
| Minimum market regimes Runs by hand onlyA timing score is only shown when those scored changes fall in at least 2 different market regimes.A regime pairs BTC's 30-day direction (up or down) with its 30-day volatility (low, middle or high third), giving 6 in all. | 2 • | Measures | By hand | Starting value first guess |
| Funding-adjusted profit and drawdown Runs by hand onlyLong-term wallets are measured on closed-trade profit plus funding received or paid. They are also measured on the largest drop in that running total, as a share of their median account value.This covers only the stretch of trade history the exchange still serves for the wallet. | — | Measures | By hand | Starting value |
| Ranking without a follow signalThe long-term tab lists scored wallets by timing score, highest first with unscored ones last, then by smallest funding-adjusted drawdown. That list is never ordered by profit, and it carries no rank, verdict or tier, so nothing on it is a follow signal.Spot holders not yet scored are listed underneath by all-time profit, up to 300. The Inspector's long-term filter and the Terminal's long-term panel still sort these wallets by all-time profit, and a wallet re-scored on several days appears once per day. | — | Shows | Every page load | Starting value |
| Scoring batch Runs by hand onlyEach scoring run takes up to 50 long-term wallets, with a budget of 300 requests to the exchange.Runs do not skip wallets that are already scored, so repeat runs tend to re-score the same ones. | 50 | Schedule | By hand | Starting value |
| Timing score reads at most 12 coins Runs by hand onlyThe timing score uses price history for at most 12 of a wallet's coins: the first 12 in alphabetical order, not the largest. Exposure changes in any other coin are left out. | 12 | Measures | By hand | Starting value |
| Exposure changes capped at 3x Runs by hand onlyEach exposure change is capped at 3 times the position held at the start of that day, up or down, so one huge add cannot dominate the score. A change from a flat start counts as exactly 1. | 3 | Measures | By hand | Starting value |
Labels that never remove anyone Marks shown on a row, with no effect on the screen. · 12 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Account size bandEvery wallet on the Hyperliquid leaderboard is tagged by account size: under $100K, $100K to $500K, $500K to $1M, or $1M and up.This is a label only; the $1M account minimum is a separate rule. | $100K / $500K / $1M | Labels | Every 48 h | Owner's decision |
| Active this weekA wallet is tagged active 7d when the leaderboard shows any trading volume for it in the last 7 days. | — | Labels | Every 48 h | Owner's decision |
| Active this monthA wallet is tagged active 30d when the leaderboard shows any trading volume for it in the last 30 days.Where the 90-day activity check has no answer yet, this tag stands in for it. | — | Labels | Every 48 h | Owner's decision |
| Losing weekA wallet is tagged red week when it traded in the last 7 days and ended them with a loss.This is a label only; it never removes anyone. | — | Labels | Every 48 h | Owner's decision |
| Losing month Not enforcedA wallet is tagged red month when it traded in the last 30 days and ended them with a loss.It is recorded but not shown anywhere on the site. | — | Labels | Every 48 h | Owner's decision |
| Low return Not enforcedA wallet is tagged low return when its all-time return on the leaderboard is under 50%. The figure is not adjusted for deposits.The tag is recorded but not shown; the site shows the return figure itself. | 50% • | Labels | Every 48 h | Starting value |
| Market-maker-like labelA wallet is labelled market-maker-like when its lifetime profit is less than 5 basis points (0.05%) of its lifetime trading volume. The label removes no one.A starting guess, not yet calibrated. | 5 bps • | Labels | Every 48 h | Starting value first guess |
| Month consistency labels Runs by hand onlyWhen the optional portfolio pass is run, each wallet also gets labels from its last month of perp trading: the share of its active days that were winners, its best day's share of the month's profit (only when the month was profitable), and its perp profit for the month and the week. These are labels only.No scheduled job runs this pass, so the positive-day share, best-day share and perp week profit on file come from earlier manual runs; the perp month profit of a perp trader that passed the screen is overwritten daily by the perp P/L job. | — | Labels | By hand | Starting value |
| Activity probe Runs by hand onlyA probe, run by hand, records whether each wallet waiting for the deep check traded in the last 90 days and, whenever it did, the date of its newest trade. A wallet with a trade on record inside the window needs no request; every other wallet's fills are read. Run on screen passers instead, it re-reads the fills of every passer, whatever its state, with no trade on record in the last 90 days. It never removes anyone.Every wallet it marks active gets a date, and pages count the days from that date to today. Its 12 Sep 2026 run marked wallets the leaderboard showed active at 30 days without asking and without a date; until another job or a re-run of the probe records a date for them, they show no last trade and are left out of the 'Traded within' filter, and of the top-trader lists unless they hold a running trade. | 90 days | Labels | By hand | Owner's decision |
| One-off: perp trader vs mixed RetiredIn a one-off run on 12 September 2026, each wallet with $500K+ of perp profit was labelled perp trader if perps made at least half its lifetime profit, and mixed otherwise. The label itself removed no one. | 50% | Labels | Once, by hand (12 Sep 2026) | Owner's decision |
| One-off: staked share RetiredIn a one-off run on 12 September 2026, each wallet with $500K+ of perp profit had its staked HYPE stored as a share of its account value plus that stake, priced at that day's HYPE price. It was a label only and removed no one.The deep check applies its own, separate staking rule. | — | Labels | Once, by hand (12 Sep 2026) | Owner's decision |
| Trading style labelA trader's profile is labelled Scalp when its median trade is held under 15 minutes, Intraday under 6 hours, Swing under 14 days, and Position at 14 days or more. The label never removes anyone.On Hyperliquid it is shown only for wallets that have had the deep check; Aster, Pacifica and Lighter profiles show it from their own trades. The stricter scalper rule (median hold under 1 hour) is separate. | 15 min / 6 h / 14 days | Labels | Every page load | Starting value |
How the data is handled History limits, equity maths, what is stored. · 35 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Every screen passer's trades, as round tripsEvery wallet that passed the screen, including one a later check rejected, has its trades read from Hyperliquid as far back as it still serves them and rebuilt into round trips, each with when and at what average price it went in and out, its largest size, fees and profit. The round trips are kept; the trades themselves are not. Each later read takes only the trades since the last one: a round trip still open carries on in the same row, so no round trip is stored twice and nothing already stored is lost.Hyperliquid serves a busy account only its last 10,000 or so trades, which for the busiest is hours; a position already open where that history starts is marked censored, its entry unknown, and a position that closed while no trades were served is closed and marked the same way. The first pass over every passer takes a day or two. | every 30 minutes | Measures | Every 30 minutes | Owner's decision |
| A wallet's trades are read again every 6 or 24 hoursA wallet whose newest trade is under 7 days old has its new trades read every 6 hours; any other wallet once a day; a wallet Hyperliquid serves only hours of history is read again within half of that, but not more often than every 30 minutes (until 17 Oct 2026 every 2 hours, which left gaps in the busiest wallets' trades). Wallets due again are read before new ones. A run stops at 300 look-ups or 15 minutes, so the 30-minute position sweep keeps its turn. | 6 hours • | Measures | Every 30 minutes | Starting value |
| Profit by coin since 22 Mar 2025 is read from Hypedexer Runs by hand onlyFor the screen passers, each trader's fills are also read from Hypedexer's index of every Hyperliquid fill as one total per day and coin -- the exchange's own closed profit, fees, size traded and number of fills -- back to 22 Mar 2025, which Hyperliquid itself no longer serves for a busy account. A free key has 5,000 credits a month (about 2 a call and 0.1 a row), so the reading works to a budget: days since a wallet's last read first, then wallets never read, biggest profit first, then older stretches; it stops at its credit cap for the run (300), at 400 for one wallet, and never spends the account below 500. It is run by hand until it is scheduled.Checked against fills read straight from Hyperliquid for one wallet (0x888e, 29,299 fills): closed profit identical to the cent over 27,556 of the 27,558 fills compared (two missing, on one coin), fees within $2. One wallet is a small sample. Hypedexer's own profit figures (coin totals, completed trades) are not used: they differed by $1.07M on that wallet. Fees are not indexed before 25 May 2025; spot coins are left out. | 22 Mar 2025 | Measures | By hand | Starting value |
| Refresh overwrites only leaderboard numbersEach census refresh overwrites only the leaderboard's own numbers stored for a wallet (account value, profit, return, volume and the quick labels). Screen results such as the pass mark and trader type, and everything the deeper checks found, are kept.A wallet the screen removed for account size or inactivity also keeps its reason and re-check date. | 17 fields | Shows | Every 48 h | Fixed |
| Tape: each trade counted once Runs by hand onlyThe tape harvester remembers the last 500,000 trade ids, so a reconnect never counts a trade twice.The tape harvester is not currently running. | 500,000 • | Measures | By hand | Fixed |
| One job at a timeThe background runner looks for due jobs a minute after its last check or job finished, and starts at most one at a time, taking the job whose last success is oldest. Its jobs therefore never run at the same time.A job is due once its full period has passed since its last success. | 60 s | Schedule | Every minute | Starting value |
| Retry a failed jobA job that fails, such as a broken leaderboard download, is tried again 15 minutes after the failed run started instead of waiting for its full period.A run that failed after 15 minutes or more, including one stopped at the one-hour limit, is tried again at the next check. | 900 s • | Schedule | Every minute | Starting value |
| One-hour job limitA job still running after one hour is stopped and counted as failed.A timed-out job is tried again at the next check, about a minute later: the 15-minute retry wait counts from when the failed run started. | 1 h | Schedule | Every minute | Starting value |
| Pause when the disk is nearly fullWhen the server has less than 8% of its disk free, the background runner starts no job that minute, not even the census or the screen.Nothing is skipped for good: due jobs start once there is room again. | 8% • | Schedule | Every minute | Starting value |
| Resuming from stored fills Runs by hand onlyWhen the deep check revisits a wallet that already has stored fills, it pulls fills only from one day before the last stored one and adds the stored fills back before rebuilding.The by-hand crawler and the by-hand monitor both do this. Funding is read over the last 30 days, not only over the new fills. | 1 day • | Measures | By hand | Starting value |
| Exchange history capThe exchange serves full fill history only for accounts with up to about 10,000 lifetime fills. Busier accounts get roughly the last 50 to 80 days, and nothing older is used.Older fills would need Hyperliquid's own archive of fills (not set up) or a paid data vendor (not used). | ~10,000 | Measures | Every fill pull | Fixed |
| Exchange page sizesThe exchange returns at most 2,000 fills, 2,000 TWAP slices, 2,000 ledger rows or 500 funding rows per request, so each history is read page by page until a short page comes back. | 2,000 • | Measures | Every fill pull | Fixed |
| Ledger and funding read limits Runs by hand onlyThe deep check reads at most 100,000 ledger rows and 100,000 funding rows per wallet, and any newer rows past that are silently left out.Unlike a cut-off fill history, this is not flagged and does not remove the wallet. | 100,000 | Measures | By hand | Starting value |
| Request budgetEach program stays within 1,100 of the exchange's 1,200 request-weight per minute, and backs off after a rate-limit reply.Each request is tried up to 6 times, and a rate-limit reply pauses further requests for 10 to 60 seconds. | 1,100 | Measures | Every request to Hyperliquid | Starting value |
| Raw fills kept only for survivors Runs by hand onlyRaw fills are stored only for wallets the deep check leaves as Candidate, Proven, Too young, Long-term or Demoted. A wallet the deep check removes keeps its own row, with its reason and the figures behind it, and no rebuilt trades or raw fills.A wallet still under evaluation keeps its rebuilt trades and figures but not its raw fills; if it passed the screen, the 3-hourly equity job still stores an equity history for it. | — | Measures | By hand | Starting value |
| Equity curve sourceEquity curves come from the exchange's own account-value history, which keeps about 100 points per account (roughly weekly for a two-year account), reduced to one point per UTC day. The deep check uses the perp account's history; the regular equity job stores the whole account's. | ~100 | Measures | Every 3 h (equity job); by hand in the deep check | Fixed |
| Where deposits come fromDeposits and withdrawals in each interval are taken from the exchange's own accounting: the change in account value minus the change in profit. The transfer ledger is used only when that figure is missing.The ledger alone misses some moves between the perp and spot sides of an account. | — | Measures | Every 3 h (equity job); by hand in the deep check, which alone falls back to the ledger | Starting value |
| Whole-account fallback Runs by hand onlyIf the transfer ledger leaves more than 20% of the perp curve's deposits and withdrawals unexplained, the whole-account curve (perp plus spot) is used instead, when the ledger explains it better.A starting guess, not yet calibrated. | 20% • | Measures | By hand | Starting value first guess |
| Return per interval Runs by hand onlyEach interval's return is its profit divided by its starting value plus half its net deposits. The returns are chained into a time-weighted return, and Sharpe and volatility are annualised using the real interval length. | — | Measures | By hand | Starting value |
| Tiny starting balances skipped Runs by hand onlyAn interval is left out if it starts from less than $100, or from less than 1% of the money deposited during it.It is also left out when withdrawals during it reach twice its starting value. | $100 • | Measures | By hand | Starting value |
| Loss floor per interval Runs by hand onlyNo interval is counted as losing more than 99%. An interval that would is capped there, and the wallet's whole return series is then marked unreliable. | -99% • | Measures | By hand | Starting value |
| Conduit account (typical flow) Runs by hand onlyIf, in a typical interval, net deposits or withdrawals exceed 50% of the starting value, the account is treated as a pass-through and its time-weighted return is marked unreliable.A starting guess, not yet calibrated; the forward test uses the same limit. | 50% • | Measures | By hand | Starting value first guess |
| Conduit account (single flow) Runs by hand onlyThe time-weighted return is also marked unreliable if any single interval's net deposits or withdrawals exceed 3 times its starting value.A starting guess, not yet calibrated. | 3 • | Measures | By hand | Starting value first guess |
| Fallback when returns are unreliable Runs by hand onlyWhen the time-weighted return is unreliable it is not reported. Sharpe, drawdown, edge decay, volatility and the rolling 30-day test are then worked out from the exchange's cumulative profit instead, which deposits cannot distort. | — | Measures | By hand | Starting value |
| Profit drawdown Runs by hand onlyProfit drawdown is the fall from the highest cumulative profit, divided by the median account value since that high. Above 100% means the trader lost more than the account it typically ran.Used by the 35% maximum-drawdown check when returns are unreliable. | — | Measures | By hand | Starting value |
| Minimum data for Sharpe Runs by hand onlySharpe is worked out only from at least 10 intervals, and is annualised using the actual average interval length.The t-statistic is Sharpe times the square root of the number of years covered. | 10 | Measures | By hand | Starting value |
| Minimum data for edge decay Runs by hand onlyEdge decay (second-half Sharpe divided by first-half Sharpe) is worked out only from at least 40 intervals (40 profit readings when returns are unreliable), and only when the first half's Sharpe is at least 0.25.Otherwise it is left blank and the edge-decay check does not apply. | 40 | Measures | By hand | Starting value |
| Rolling window length Runs by hand onlyThe rolling test checks, at each point of the curve, whether the trailing 30-day return was positive, or the trailing 30-day profit when returns are unreliable.Feeds the Proven-only check that at least 55% of those 30-day windows are positive. | 30 days | Measures | By hand | Starting value |
| New equity curves fetched every 3 hoursEvery 3 hours, screen-passing wallets with no stored equity history get one from the exchange, one request each and at most 400 a run. A curve already stored is not refreshed by this job.So a profile's equity curve stays as first fetched until a deep check rewrites it. | 3 h; <= 400 calls | Schedule | Every 3 h | Starting value |
| Data retentionEvery day, raw tape trades older than 14 days, position-change events older than 180 days and lifecycle changes older than one year are deleted.Removed wallets' rows, rebuilt trades, equity history and forward-test records are never deleted. | tape 14 d; events 180 d; lifecycle 365 d | Schedule | Daily and nightly at 04:20 | Starting value |
| Daily position readings kept 8 daysThe daily readings of every account's open positions are kept 8 days: today back to 7 days ago and the day before that, for an account not read that day. Every older day is deleted.A position's 3D OI needs only 4 of those days, but the 7D totals are added up from each account's own reading, so the readings stay until then. | 8 days • | Schedule | Daily and nightly at 04:20 | Starting value |
| Hourly position readings kept 8 daysThe hourly readings of every tracked account's open positions are kept 8 days: the 7 days a line shows and the day before them, so the first hours on a line can still stand at the reading before them. Every older day is deleted, in whole UTC days, by the nightly clean-up, like the daily readings.One reading per account, coin and hour, the hour's last, written by the same reads as the daily ones: no extra request to any venue. | 8 days • | Schedule | Daily and nightly at 04:20 | Starting value |
| Dust balances count as nothingWhen working out how long a spot token has been held, a leftover balance worth under $1, or no more than 0.1% of the size of the trade next to it, counts as nothing: a buy on top of it starts a new holding, and a sell that leaves only that much ends one. | $1 • | Measures | Daily | Starting value |
| What a month is in trade tests Runs by hand onlyIn the trade-based tests (profit without the best month, share of profitable months, biggest single month) a month is a fixed 30-day block counted from 1 January 1970 by each trade's close date, not a calendar month.The monthly consistency checks switch to calendar months from the equity curve when a wallet's trade history is cut short. | 30 days | Measures | By hand | Starting value |
| Optional clean-up of raw fills Runs by hand onlyWhen the clean-up is run by hand with its fills option, raw fills are deleted for every wallet except Candidate, Proven and Demoted traders and wallets still under evaluation that changed in the last 7 days. Their rebuilt trades and figures stay.The daily and nightly clean-ups never do this. | 7 days | Schedule | By hand | Starting value |
What the site shows Who each tab lists, and how fresh the numbers are. · 96 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Profit by coin on a trader's pageWhen daily fills have been read for a wallet, its page lists every perp coin's closed profit less fees over all the days read, best first: five coins, the rest behind "show more", every column sortable, with the total and the dates it covers. Funding, spot trades and the moves of open positions are not in it.The table of round trips on record, further down, covers only the trades Hyperliquid still serves, which for a busy trader is hours. | — | Shows | Every page load | Fixed |
| Who the main tabs showOf Hyperliquid wallets, only those that passed the screen appear on the Dashboard, the Tracker's default view and the Open Interest tab. The Tracker shows where each one's profit comes from; the Open Interest tab does not show a type.The Tracker's Screened out view lists the rest with their reason, and the Whitelist shows starred wallets whatever their result. | — | Shows | Every page load | Fixed |
| Passing is permanentOnce a wallet passes the screen it stays on the Dashboard, Tracker and Open Interest tabs for good. The daily screen does not re-check it for account size or recent activity, and it stays listed even if its profit falls under $500K or a later check removes it.The Tracker, the Whitelist and the Open Interest tab show a later removal as a chip with its reason beside the wallet's name. | — | Shows | Every page load | Fixed |
| Funnel counts freshnessThe funnel counts on the Rules page are read live from the database and re-read at most once every 10 minutes.After a quiet spell, a copy of the page up to 30 minutes old can be shown once while a fresh one is drawn. | 600 s • | Shows | Every page load (kept up to a minute) | Starting value |
| Distorted-return warningA trader's profile flags its compounded return with a warning when, in any interval, net deposits or withdrawals were larger than the equity at the start of that interval, and points the reader to dollar P/L and drawdown instead.Above 3 times the equity the compounded return is withheld and only the warning shows. | 1 | Shows | Every page load | Starting value |
| Forward-test status lineThe Rules page shows how many listed traders have been frozen for the forward test, and in how many cohorts, or says it has not started. Favourites and the control group are not counted.Cohorts are frozen by hand, and a trader frozen in two cohorts counts twice. | 10 min • | Shows | Every page load (kept up to a minute) | Starting value |
| Who the Hyperliquid pages countThe Hyperliquid dashboard and the Open Interest tab count every wallet that passed the screen, including wallets a later check rejected, such as market makers and basis desks. The Open Interest tab labels such a wallet with its rejection reason; the dashboard shows it unlabelled.Rejected hedgers and market makers can therefore hold a large share of the open interest shown. | — | Shows | Every page load (kept up to a minute) | Owner's decision |
| Headline cards count perp and spot traders onlyThe Hyperliquid summary cards and the page's trader count include only screen passers typed as perp traders or spot traders. Stakers that passed the screen are left out of that count and of the cards' P/L and account value.The tracker's 'kept' total includes stakers, so it is higher. | — | Shows | Every page load (kept up to a minute) | Starting value |
| Headline open interest: perp-typed wallets read in 2 hThe Hyperliquid cards' open interest and unrealized P/L add up the positions of every wallet typed as a perp trader that was read in the last 2 hours. The spot card's value held does the same for wallets typed as spot traders.Unlike the token tables, this is not limited to wallets that passed the screen, and stakers' positions are left out. | 2 h | Shows | Every page load (kept up to a minute) | Starting value |
| Dashboard counts only rows read in the last 2 hOn the dashboard, an open position or spot balance counts toward the cohort's token tables and liquidation bands only if it was read in the last 2 hours. An older row belongs to a wallet the sweep has stopped reaching.A perp trader's or staker's spot balances are re-read only about every 2 hours, so they can drop out of the spot figures briefly between reads. | 2 h | Shows | Every page load (kept up to a minute) | Starting value |
| Perp-only P/L shown while under 2 days oldPerp-only P/L from the daily perp read is used only while it is under 2 days old. The all-venues Hyperliquid card's 'perp 30d P/L' adds up only fresh readings and says how many traders it covers when that is not all of them. The top-5 perp ranking falls back to the perp profit measured at screening when a reading is older. | 2 days • | Shows | Every page load (kept up to a minute) | Starting value |
| Live price used only if under 10 min oldA Hyperliquid position's value and unrealized P/L are re-marked at the live mid price only when that price is under 10 minutes old. Otherwise the sweep's stored value is used, and the Open Interest tab says on hover that the feed is down.The dashboard's coin totals always use the sweep's stored values. | 10 min | Shows | Every page load (kept up to a minute) | Starting value |
| Price feed shown live under 2 minThe dashboard's price-feed dot shows live when the newest stored price is under 2 minutes old, and stale otherwise. The page re-checks it every 2 seconds. | 120 s | Shows | Every page load | Starting value |
| Top 5 perp coins and top 5 spot tokensThe dashboard lists the cohort's top 5 perp coins by open interest (long plus short) and its top 5 spot tokens by value held, counting only rows read in the last 2 hours.The 'N of M traders swept' line counts traders holding something, not every trader read. | 5 | Shows | Every page load (kept up to a minute) | Starting value |
| Where the cohort liquidatesFor each top coin, the dashboard's Liquidate Long/Short column shows, for the longs over the shorts, the price band holding the middle half of that side's money (the notional-weighted 25th to 75th percentile of liquidation prices), and how many positions on that side report a liquidation price. Positions with no liquidation price are left out.Unlike the Open Interest tab, the Hyperliquid dashboard does not drop liquidation prices on the wrong side of the mark; the all-venues view's bands use the Open Interest tab's prices, which do. | 25th-75th percentile | Measures | Every page load (kept up to a minute) | Starting value |
| Top traders: a trade in 30 days or a running tradeEvery top-trader list (the top-5 perp and spot lists, each venue's top 5 and the all-venues view's top 3) takes accounts that traded in the last 30 days, counted from the trade's date to today, or hold a running trade now: an open position of at least $1,000, or for a spot trader also a holding of at least $1,000 that is not a stablecoin. Hyperliquid's lists take screen passers of that type, perp traders ranked on perp profit and spot traders on spot profit; the other venues rank on their board's own P/L window.Aster dates a trade from the account's last order, or from a trade in its 30-day board window when that window ended in the last 30 days (orders in privacy mode are never stored); Pacifica and Lighter publish no trade time, so only a trade in their 7-day window counts. Hyperliquid's trade date moves forward when the live fill feed, the 30-minute position sweep, the position and spot history jobs, the daily screen or the probe see a newer trade. | 30 days • | Shows | Every page load (kept up to a minute) | Owner's decision |
| A running trade counts from $1,000For the top-trader lists, an open position counts as a running trade, and so as active, when it is worth at least $1,000: on Hyperliquid read in the last 2 hours, on the other venues in the venue's latest position read. A spot trader's open perp position counts too, and so does its holding of at least $1,000 that is not a stablecoin. Smaller ones do not make a quiet account active.The rule replaced 'a trade in 90 days, or $1M+ of open interest in profit' on 16 Oct 2026. | $1,000 • | Shows | Every page load (kept up to a minute) | Starting value |
| Top traders need an account of $1,000No top-trader list shows an account worth under $1,000, as the Tracker does not, so an account that made its profit and emptied itself does not hold a place. On Hyperliquid the value is the one the census refreshes every 48 hours, or the position sweep's from the last 2 hours; an account whose value is not known stays. | $1,000 • | Shows | Every page load (kept up to a minute) | Owner's decision |
| Each top trader's book: 3 shown, all on a clickEach top trader shows its 3 biggest open positions or holdings; a click on its bar shows every one, biggest first. The bar also adds up the unrealized P/L of its open positions.These books show the latest stored read, even when it is older than 2 hours. | 3 (all on a click) | Shows | Every page load (kept up to a minute) | Owner's decision |
| All-venues view: top 10 coins, top 3 traders per venueThe all-venues view adds the four venues' books together by coin and lists the 10 coins with the most open interest, each venue valued at its own prices: each venue's open interest now beside its long and short open interest 3 days ago (top) and 7 days ago (below), then the average entry of the longs and the shorts and where each side liquidates, over every venue's positions together. It also lists the top 3 traders per venue: on Hyperliquid by perp P/L, on the others by their board's own P/L window, by the same rule as every top-trader list.It reads the Open Interest tab's book, so a thousand-token contract such as Lighter's 1000PEPE is summed with kPEPE, and Aster's positions, which carry no liquidation price, are not in the liquidation bands. | 10 coins; 3 traders per venue | Shows | Every page load (kept up to a minute) | Starting value |
| Pages open on HyperliquidA page with no venue chosen, or an unknown one, opens on Hyperliquid. 'All venues' is one menu click away.Hyperliquid is the only venue the screen runs on. | hyperliquid • | Shows | Every page load | Owner's decision |
| Open Interest tab: Hyperliquid rows read in the last 2 hOn the Open Interest tab, a Hyperliquid position or spot holding counts only if it was read in the last 2 hours.A perp trader's or staker's spot balances are re-read only about every 2 hours, so they can drop out briefly between reads. | 2 h • | Shows | Every page load (kept up to a minute) | Starting value |
| Spot holding must be in the wallet's newest readA Hyperliquid spot holding counts only if it came in the wallet's newest read (within 60 seconds of it). A token missing from the newest read is treated as sold. | 60 s • | Shows | Every page load (kept up to a minute) | Starting value |
| Lighter and Aster holdings older than 6 h left outLighter and Aster spot holdings not re-read in the last 6 hours are left out, and the page says how many. | 6 h • | Shows | Every page load (kept up to a minute) | Starting value |
| History figures used only if read in 36 hA position's open date and realised-since-open are shown only if their history read is under 36 hours old. The same applies to a spot holding's held-since date, realised profit, 7-day buys and sells, and stablecoin conversions; otherwise the cell shows a dash with the reason. | 36 h • | Shows | Every page load (kept up to a minute) | Starting value |
| Position changed by a live fill: size onlyIf a live fill changed a Hyperliquid position after the last sweep, only the new size and notional are shown. Entry, liquidation price and unrealized P/L stay blank until the next sweep.Only about 15 of the largest wallets are followed fill by fill; the rest change only when the sweep reads them. | 2 s • | Shows | Every page load (kept up to a minute) | Starting value |
| Open date must belong to this positionA stored open date and realised P/L are attached to a position only if they are for the same side. On Hyperliquid they must also have been read after the position was first seen; on Pacifica the stored open time must match the position's own within 60 seconds. | 60 s | Shows | Every page load (kept up to a minute) | Starting value |
| Liquidation price must be reachableOn the Open Interest tab, a liquidation price is shown only if it is above zero and on the losing side of the mark: below it for a long, above it for a short. Otherwise the cell shows a dash with the reason on hover.Aster publishes no liquidation prices, and the dashboard's bands skip only zero or negative prices. | — | Shows | Every page load (kept up to a minute) | Starting value |
| Lighter/Pacifica liquidation prices over 10x from the mark droppedOn Lighter and Pacifica, a liquidation price more than 10 times the mark (or the entry, when there is no mark), or under a tenth of it, is dropped as an account-level figure rather than this position's.Hyperliquid prices are exempt because a low-leverage long can really liquidate that far below the mark. | 10x • | Shows | Every page load (kept up to a minute) | Starting value |
| Leverage shown is the setting; effective leverage on hoverThe leverage shown for a position is the venue's setting, the most the account allows, not what it uses. Hovering shows the effective leverage: the position's notional divided by the account value.A coin's 'Avg leverage' is the notional-weighted average of these settings. | — | Measures | Every page load (kept up to a minute) | Starting value |
| Stablecoins are a named listStablecoins are a fixed list of named tokens. No token is treated as a dollar just because 'USD' appears in its name.Only these tokens get the stablecoin 7-day change and conversion figures. | USDC, USDT0, USDE, USDH, USDHL, FEUSD, USDXL, USDG, USDT, USD1, USDF, BUSD, FDUSD • | Labels | Every page load | Starting value |
| Stablecoins valued at $1On the Open Interest tab, a stablecoin with no stored price is valued at $1. The Hyperliquid sweep always prices Hyperliquid's own stablecoins at $1, never at a thin order book's mid. | $1 | Measures | Every page load | Starting value |
| Hyperliquid spot priced from spot pairsHyperliquid spot holdings are valued at a USD price from the spot pairs quoted in a stablecoin (the USDC pair first), never from perp prices, because the same name can be two different assets.A token with no stablecoin-quoted pair gets no value. | — | Measures | Every 30 min | Starting value |
| 7-day balance changeA holding's 7-day change is today's amount minus the address's reading on, or nearest before, the day 7 days ago. With no reading that old, the page says 'tracking since' and a date instead of a number.A token not held on that day counts from zero, and token totals show this change for stablecoins only. | 7 days | Measures | Every page load (kept up to a minute) | Starting value |
| 3D and 7D: open interest thenThe Dashboard shows the cohort's Net OI per coin 3 and 7 days back (Net 3D, Net 7D) and the Open Interest tab its long and short open interest then (3D Long/Short, 7D Long/Short), on every venue; the Tracker shows each trader's open interest 3 days back (3D OI). Every table that lists a trader's positions (the Dashboard's top 5 traders, the Tracker's opened book, the Open Interest drill-down and every trader page) shows each position's size 3 days back (3D OI), marked added, reduced, new or flipped. They come from one reading of every account's open positions per UTC day, the day's last. Until a venue has readings that old, the page shows a dash and the date tracking started.Sizes are exact, so they show whether a trader added or reduced. Values are at that day's price (Aster's days filled in from stored snapshots at entry price), so open interest moves with the price as well as with trades. A coin's total is summed over that day's reading of the accounts shown now, so a position closed since still counts: on Hyperliquid the wallets read in the last 2 hours (the ones Net OI counts), on Lighter and Pacifica the board's top 40 of the latest hour, on Aster every board trader. A coin Hyperliquid's sweep does not read (builder markets, spot) shows a dash. | 3 days; 7 days | Measures | Every page load (kept up to five minutes) | Starting value |
| An account not read that day uses the day beforeWhen an account had no position reading on the day 3 or 7 days back, its reading from the day before is used; with neither, its value is a dash.Hyperliquid is read every 30 minutes, Lighter and Pacifica hourly for the board's top 40, Aster every 2 hours. The hover text says when a reading is from the day before. | 1 day • | Measures | Every page load (kept up to five minutes) | Starting value |
| Added or reduced sinceIn a position's 3D OI cell (its size 3 days back), an up arrow means it is bigger now and a down arrow smaller, when the size moved by more than 0.5%. Flat then reads 'new'; the other side then reads 'flipped'. | 0.5% • | Shows | Every page load (kept up to five minutes) | Starting value |
| 7-day open interest line, one reading an hourThe perp tables on the Dashboard, the Tracker and the Open Interest tab, and the rail of every trader page, draw the tracked traders' open interest over the last 7 days as a line, one reading an hour (the hour's last). A coin row draws the long and the short open interest of the traders counted now, each trader counted from its first hourly reading; a trader row or page draws that trader's long and short; a position draws its size, green while long and red while short. Hovering the line gives any hour's values.Hyperliquid is read every 30 minutes, Lighter and Pacifica hourly (the board's top 40), Aster every 2 hours; an hour with no reading of its own stands at the account's reading from up to 3 hours before. The hourly readings started on 13 Oct 2026; until 7 days of them exist, a line covers the hours read so far. | 7 days • | Shows | Every page load (kept up to five minutes); on a Lighter, Pacifica or Aster trader page, as old as the stored page, rebuilt every few hours | Owner's decision |
| An hour missing too much open interest is left outA coin's line is drawn only for the hours whose readings hold at least 97% of the counted traders' open interest, a trader unread in that hour valued at its nearest reading. Otherwise the hour is left out and the line breaks there, so a sweep that fell short never reads as traders closing.A trader counts toward the hours from its first hourly reading in the 7 days, so one that joined the group mid-week (a new name in a venue's top 40) adds its positions from then on and does not empty the hours before it. | 97% • | Shows | Every page load (kept up to five minutes) | Starting value |
| An hour with no reading stands at the one before, up to 3 hoursIn an hour with no reading of its own, an account stands at its latest reading from up to 3 hours before, as the live figures count a position read in the last 2 hours. Older than that, the account is unread in that hour: a coin's line leaves the hour out if that matters (see the 97% rule), a position's or trader's line breaks there.Aster is read every 2 hours, so every other hour of an Aster line stands at the hour before; a sweep still running in the current hour stands its unread wallets at their previous hour. Nothing is filled into the stored readings: each page works this out as it reads them. | 3 hours • | Shows | Every page load (kept up to five minutes) | Starting value |
| Venues summed: a short hour stands at the nearest oneWhere venues are added together (the all-venues view and the Open Interest tab across venues), a coin's 7-day line is drawn in every hour at least one venue holding it has enough readings for. A venue short of readings in that hour adds what it held in its nearest hour that has them, the hour before or else the first one after, so the summed line does not break each time one venue misses an hour.Lighter and Pacifica follow their board's top 40, which changes hour to hour, so on their own they miss many hours; their own lines still break there. | — | Shows | Every page load (kept up to five minutes) | Owner's decision |
| OI High/Low and Net OI High/Low: the highest and lowest hour of 7 daysA position's OI High/Low shows its biggest and its smallest hourly size of the last 7 days, on either side. A coin row's Net OI High/Low shows the highest and the lowest hourly net open interest (long minus short) of the traders counted now, green while net long and red while net short. Hovering gives the hours they were read in.Until 7 days of hourly readings exist, it is over the hours read so far, and the hover says from when. | 7 days • | Shows | Every page load (kept up to five minutes) | Owner's decision |
| Stablecoin 7-day change includes holders who sold outThe cohort's 7-day stablecoin change adds up every address read both a week ago and since, including holders who have since sold out, not just current holders.In the all-venues view Aster's stablecoins are left out, because they are its perp margin. | — | Measures | Every page load (kept up to a minute) | Starting value |
| One coin name across venues (Open Interest tab)On the Open Interest tab, coin names are made the same across venues so one row adds up one asset: Aster's quote suffix is dropped (BTCUSDT becomes BTC), and seven thousand-token contracts named 1000X become kX.The dashboard's all-venues table drops Aster's suffix but keeps 1000X names. | PEPE, SHIB, BONK, LUNC, FLOKI, DOGS, NEIRO • | Measures | Every page load | Starting value |
| A coin opens into every position in itOpening a coin or token on the Open Interest tab shows every position or holder in it, biggest first. A long list scrolls inside its own box. | all | Shows | On click | Owner's decision |
| Open Interest tab books rebuilt at most once a minuteThe Open Interest tab's books are rebuilt at most once a minute and shared by every sort, filter and drill-down. | 60 s • | Shows | Every page load | Starting value |
| Tracker opens on wallets the screen keptThe tracker opens on the wallets the screen kept. Ticking 'Whitelist only' shows every starred wallet with a leaderboard record, kept or not.'Kept' includes wallets a later check rejected. | kept by the screen | Shows | Every page load (kept up to a minute) | Starting value |
| What 'everything we hold' means'Everything we hold' in the tracker is every wallet with a leaderboard all-time P/L on record, including those under the $500K floor. 'Screened out' is every such wallet without a screen pass.'Screened out' also includes $500K+ wallets the screen has not reached yet. | — | Shows | Every page load (kept up to a minute) | Starting value |
| Tracker 'Traded within' filterFor Hyperliquid wallets, the 'Traded within' filter keeps wallets whose last trade on record is at most 1, 7, 30 or 90 days before now, counted from the trade's date.A wallet leaves a window as time passes unless a newer trade is recorded, and a wallet with no date is left out of every window. The live fill feed moves a date at most once an hour; a position change seen by the 30-minute sweep is dated to the sweep's previous read. | 24 hours / 7 days / 30 days / 90 days / any time • | Shows | Every page load (kept up to a minute) | Starting value |
| Last trade marked fresh or staleThe 'Last trade' cell is marked fresh at 7 days or less and stale beyond 60 days.For Hyperliquid wallets the days are counted from the last trade on record to now. | fresh <= 7 d, stale > 60 d | Labels | Every page load (kept up to a minute) | Starting value |
| The chip beside a wallet's nameThe Tracker and Whitelist have no screen-result column. A chip beside the name shows a later verdict (Proven, Candidate, Slipping, Long-term), the rule that removed the wallet -- even when it passed the screen and stays listed -- 'Not screened yet' for a wallet the daily screen has not reached, or 'Screened out'.Open Interest tab rows carry the same reason chip. | — | Labels | Every page load (kept up to a minute) | Starting value |
| Type column: every real source of profitThe Tracker and Whitelist name each profit source -- perps, spot, staking -- that is at least 20% of the wallet's winning profit, always in that order, so a wallet reads Perp, Spot, Staking or a combination such as Perp/Spot; the split is shown on hover.When the losing sources outweigh the winning ones, or nothing made money, the screen's single type is shown instead. Rows from other venues split perps and spot only. | 20% • | Shows | Every page load (kept up to a minute) | Starting value |
| Open interest against the account (OI/Acc)Each position in a coin's drill-down shows the trader's account value and, as OI/Acc, the position as a multiple of it (notional divided by account value). Each spot holding shows the holder's spot book -- everything it holds in spot on that venue, stablecoins included -- and the holding's share of it.For Hyperliquid perps the account is the perp account (margin plus unrealized P/L) read with the position. The spot book never adds the perp account: a unified account keeps its perp collateral in spot already. Holdings with no USD price are left out of the book, and the shares are then an upper bound. | — | Shows | Every page load (kept up to a minute) | Fixed |
| Tracker page size and default sortThe tracker shows 50 rows a page by default, or 100 or 200 when the viewer picks that beside the pager, with blanks sorted last in both directions. A click on a column title sorts by it, and a second click flips the order. The default sort is all-time P/L on Hyperliquid and Pacifica, 30-day P/L on Aster, and 7-day P/L on Lighter and on all venues. | 50 • | Shows | Every page load (kept up to a minute) | Owner's decision |
| Tracker leaves out accounts under $1,000The Tracker does not list an account worth under $1,000, on any venue, so the accounts that matter are not buried. A starred account is listed whatever its size, and so is one whose value is not known.Nothing is deleted: the account and its screen result stay on record, and the Whitelist still shows it if starred. | $1,000 • | Shows | Every page load | Owner's decision |
| Recent trades: every round trip closed in 30 days, and every open oneA trader page's Recent trades lists every round trip that closed in the last 30 days (the newest 1,000 when there are more) and every position still open, whatever its age: open ones first, marked open, with no exit, how long they have been held (a floor when the entry is unknown) and what they have made so far, the parts closed plus the live unrealised figure; then the closed ones, newest first; 10 at a time with the rest behind 'show more'. Trade history below it lists every round trip on record, open and closed (the newest 2,000), 10 a page by default or 50, 100 or 200, and either table sorts by any column when its title is clicked. Above it, the page says when the trades on record start and sets what they add up to beside Hyperliquid's own all-time perp P/L for the account. A month only partly covered by the trades on record takes its profit from the account's equity instead.Profit per round trip is before fees; fees have their own column. Hyperliquid serves an account only about its last 10,000 fills, so for a busy trader most of the all-time profit was made before the first trade on record. | 30 days • | Shows | Every page load | Owner's decision |
| Top 8 reject reasons as filtersThe tracker offers the 8 most common reject reasons, counted over every wallet on file, as one-click filters.Picking one also shows wallets outside the screen. | 8 | Shows | Every page load (kept up to a minute) | Starting value |
| Whitelist = every starred accountThe whitelist lists every starred account on every venue, whether or not it passed the screen. A starred account on another venue that has dropped off that venue's board stays listed, marked as no longer read. A trader is starred from the star before its name on the Tracker, in the Dashboard's trader lists and in the Open Interest tab's drill-downs; every star for the same account flips together.Starring an account never changes whether it passes the screen. | — | Shows | Every page load (kept up to a minute) | Starting value |
| Positions re-read every 30 minutesEvery Hyperliquid wallet that passed the screen (including ones rejected later), plus every tracked and starred wallet, has its holdings re-read about every 30 minutes, least recently read first, up to 950 calls a run. Wallets the cap leaves out are read first on the next run. | 30 min; <= 950 calls | Schedule | Every 30 min | Owner's decision |
| What is read for each kind of walletPerp traders, stakers and wallets not yet classified have their perp positions read on every sweep. Spot traders have their spot balances read, plus their perp positions while they hold one, and tracked and starred wallets get both on every sweep. | every type except spot traders | Schedule | Every 30 min | Starting value |
| Perp traders' spot balances every 2 hoursFor perp traders, stakers and wallets not yet classified, spot balances are re-read every 2 hours rather than on every 30-minute sweep.Tracked and starred wallets have their spot balances read on every sweep. | 2 h • | Schedule | Every 30 min sweep, when due | Starting value |
| Trade events come from the live stream onlyThe 30-minute sweep only corrects stored positions and records no trade events. Opens, adds, reductions, closes and flips are recorded only from the live fill stream, so only the live-watched wallets have a trade-by-trade history. | — | Shows | Every 30 min | Starting value |
| Sold tokens removed and recorded as zeroWhen a fresh read no longer shows a token, the holding is removed and that day is recorded as zero, so selling out shows as a fall to zero rather than as missing data.A read that fails removes nothing. | — | Shows | Every sweep | Starting value |
| Live fills for at most 15 walletsHyperliquid streams live fills for at most 15 wallets to one internet address, so only 15 wallets are watched trade by trade. Every other wallet relies on the 30-minute sweep. | 15 • | Shows | Always on | Fixed |
| Who gets the live slotsThe live slots go to the largest wallets by account value that passed the screen and are Proven, Candidate, Demoted, Evaluating or Long-term. Rejected wallets are left out, except for the reference slots.The list is picked when the live watcher starts and does not change until it restarts. | Proven, Candidate, Demoted, Evaluating, Long-term • | Shows | At watcher start | Starting value |
| Reference wallets log only opens, closes and flipsFor the market makers and basis desks watched as a reference, only opens, closes and flips are recorded as trade events. Their adds and reductions still update the stored position. | OPEN, CLOSE, FLIP • | Shows | Always on | Starting value |
| Live price stored on a 0.05% move or every 5 minutesA live price is stored only when it has moved 0.05% or more since it was last stored, or when 5 minutes have passed.A stored price can therefore trail the market by just under 0.05% for up to 5 minutes. | 5 bps or 300 s • | Shows | Always on | Starting value |
| Prices stored only for coins someone holdsLive prices are stored only for coins held in a tracked open Hyperliquid perp position or spot balance. That list is refreshed every 3 minutes.If no holdings are known yet, every coin's price is stored. | 180 s • | Shows | Always on | Starting value |
| Perp-only P/L re-read dailyOnce a day, the perp-only 30-day and all-time P/L of each Hyperliquid perp trader that passed the screen is re-read from the exchange, oldest reading first, up to 450 calls a run.Traders a run does not reach are read first the next day. | 450 calls | Schedule | Daily | Starting value |
| Spot trades read daily for current holdersOnce a day, the spot trades of every screen-passing Hyperliquid wallet whose balances were read in the last 48 hours are read, oldest first, up to 700 calls a run and 6 per wallet. They give each holding its start date, its realised P/L, and its 7-day buys, sells and stablecoin conversions.Tokens held at any time in the last 8 days are included, so a holding sold this week keeps its 7-day figures. | 48 h • | Schedule | Daily | Owner's decision |
| Start date searched only for holdings of $1,000+Older trades are paged through to find when a holding started only if the holding is worth $1,000 or more. Smaller holdings are dated from the trades already read, so their start date is more often a lower bound. | $1,000 • | Measures | Daily | Starting value |
| 7-day spot figures only when a full week is readA holding's 7-day buys, sells and stablecoin conversions stay blank unless the trades read cover the whole 7 days.Buys and sells also stay blank for a token with no market of its own, such as USDC. | 7 days | Measures | Daily | Starting value |
| Position open dates refreshed every 3 hoursEvery 3 hours, the open date and realised P/L since opening are worked out from trades for each open position of a screen-passing Hyperliquid wallet. Undated, flipped and reopened positions go first, then any not refreshed in 24 hours, up to 300 calls a run.The Open Interest tab ignores an open date that has not been refreshed for more than 36 hours, which is why each position is re-read at least every 24 hours. | 300 calls | Schedule | About every 3 h | Owner's decision |
| Independent price check every 30 minutes Not enforcedEvery 30 minutes, an independent price from DefiLlama is fetched for each coin held in an open perp position and compared with the stored live price. Coins that are 2% or more apart are listed in an internal log.Nothing on the site is changed or flagged when the two prices disagree. | 2% • | Measures | Every 30 min; venue map daily | Starting value |
| Funding rates collected every 30 minutes Not enforcedFunding rates on six venues (Hyperliquid, Lighter, Aster, Pacifica, Binance and Bybit) are collected every 30 minutes and converted to an annual rate.No public page shows them yet. | 30 min | Schedule | Every 30 min | Starting value |
| Venue trader pages rebuilt ahead of timeEvery 2 hours, up to 25 trader pages per venue on Aster, Pacifica and Lighter are rebuilt in advance so they open instantly, starting with pages never built and then the oldest. A run stops starting new pages after 15 minutes. | 25 per venue | Schedule | Every 2 h | Starting value |
| Data jobs run one at a timeScheduled data jobs run one at a time, with at most one starting per minute. When several are due, the one whose last successful run is oldest goes first. | 60 s | Schedule | Continuous | Starting value |
| Failed job retried after 15 minutesA data job that fails is retried 15 minutes after the failed run started rather than at every check, so one failing job cannot hold up the others.A run that failed after 15 minutes or more is tried again at the next check. | 15 min • | Schedule | Continuous | Starting value |
| Updates pause when the disk is nearly fullWhen the server's disk has less than 8% free, no scheduled data job starts and live prices stop being stored. Most figures stop updating until space is freed, rather than risk jamming the database. | 8% free • | Schedule | Every minute, and on every price update | Starting value |
| Jobs stopped after 1 hourA data job that runs for more than an hour is stopped and counted as failed. | 1 h | Schedule | Continuous | Starting value |
| Pages served from cache for 60 secondsThe home, tracker, whitelist, Open Interest, method and rules pages are served from a stored copy for up to 60 seconds after it was built.Live prices on these pages are fetched separately and never cached. | 60 s • | Shows | Every page load | Starting value |
| Older copy served while a fresh one builds, up to 30 minutesOnce a stored copy is more than 60 seconds old, it is still served at once while one fresh copy is built in the background. Only a copy older than 30 minutes makes a visitor wait for a fresh build. | 30 min • | Shows | Every page load | Starting value |
| Main pages rebuilt every 5 minutesThe home page, the all-venues home page, the tracker and the Open Interest tab are rebuilt in the background about every 5 minutes.The whitelist, method and rules pages are built only when someone visits them. | 5 min • | Schedule | Every 5 min | Starting value |
| A star or rename clears the cacheStarring or renaming a wallet empties the page cache, so the change shows on the next page load. | — | Shows | Whenever a star or name is saved | Starting value |
| A later rejection shown against an earlier passIf the full trade review passed a trader or routed it to long-term, but its latest status change moved it from Long-term, Candidate or Proven to Rejected, the profile says so and gives the reason and date. Without this, the profile would show two verdicts that seem to contradict each other. | — | Shows | Every page load | Starting value |
| Fall from peak: trading or cashing outWhen an account's value is below its peak, the profile labels the fall 'cashing out' if trading made money since the peak, 'trading loss' if no net money was taken out, and 'both' otherwise. It uses the exchange's own accounting, in which every change in account value is trading P/L plus net deposits.When both causes contributed, the page shows how much of the fall each one explains. | — | Shows | Every page load | Starting value |
| Starting capitalStarting capital is the account value on the first recorded day, before that day's deposits and withdrawals. If that comes out at zero or less, that day's deposit is used instead when there is one, and capital figures count only days with a recorded account value.Deposits and withdrawals are worked out from the exchange's own figures (change in account value minus change in P/L), not from the transfer history. | — | Shows | Every page load | Starting value |
| Monthly record: 3 to 5 yearsThe monthly record shows calendar months from the first month with data, never fewer than the last 3 years and never more than the last 5. A month's figure is the P/L of round trips closed that month, or failing that the change in the account's P/L curve; a month with neither is grey (no data), never zero.Months taken from round trips are before fees and funding, while months from the P/L curve include fees, funding and open positions. | 36 • | Shows | Every page load | Starting value |
| Avg Long/Short: coin entry prices weighted by sizeA coin's average entry price on the dashboard is weighted by position size and shown separately for longs and shorts, in one Avg Long/Short cell: the longs' price over the shorts'. A small position therefore moves it less than a large one.On a venue's dashboard it covers that venue's tracked accounts (on Hyperliquid, screen-passing wallets whose positions were read in the last 2 hours); on the all-venues view, every venue's positions together. | — | Shows | Every page load (kept up to a minute) | Starting value |
| Live slots for market makers and basis desksUp to a quarter of the live slots (3 of 15) go to the largest rejected market makers and basis desks, as a reference for where liquidity and carry sit. They take the places of the 3 smallest other wallets. | 3 | Shows | Continuous; slots are chosen once at start | Starting value |
| Stablecoin swaps are not flowsA swap between two stablecoins, for example USDT0 for USDC, is not counted as money moving into or out of stablecoins in the 7-day conversion figures.The swap still counts in the 7-day buys and sells of the stablecoin that is the pair's base (USDT0 in a USDT0/USDC swap), not in the other one's. | — | Shows | Daily | Starting value |
| Spot 'held since' dateA spot holding's 'held since' date is the last buy made from an empty balance. When that buy is not in the trades read, the earliest trade seen is shown as a lower bound ('≥ N days'); stablecoins get no date, and neither does a holding refilled by transfer after a sell-out with no trade since.A holding emptied and refilled by transfers between two trades reads as one continuous holding. | — | Shows | Daily | Starting value |
| Perp position open dateA perp position's open date is the last trade that opened it from flat or flipped it to its current side, so a position closed and reopened dates from the reopen. If the available trades do not reach that opening trade, or some are missing after it, the date is a lower bound shown as '≥ N days'.Hyperliquid serves only about the last 50 to 80 days of trades for busy accounts, so older positions show a lower bound. | — | Shows | Every 3 h | Starting value |
| Venue trader pages older than 6 hours say soAster, Pacifica and Lighter trader pages are served from a stored copy. A copy older than 6 hours carries a 'stored page, N h old' banner with a rebuild link, and only a page never built before is fetched live when someone visits. | 6 • | Shows | Every page load | Starting value |
| Playbook needs 12 closed tradesThe winners-versus-losers playbook says nothing about a trader with fewer than 12 closed round trips in the chosen window (30 days by default, or 90 days, or all time). Round trips that opened before the available trade history are not counted.The playbook page is not linked from the site yet. | 12 • | Shows | Every page load | Starting value |
| Playbook calls out a trait at 1.5xThe playbook points out a difference between winners and losers only when each side has at least 4 trades and one side's median is at least 1.5 times the other's, or one side never does it at all.The playbook page is not linked from the site yet. | 1.5 • | Shows | Every page load | Starting value |
Aster, Pacifica and Lighter Listed from each venue's own leaderboard, not screened. · 33 rules
| Rule | Value | Effect | Runs | Set as |
|---|---|---|---|---|
| Each venue ranked on its own board windowTraders on each other venue are ranked on that venue's own leaderboard window: 30 days on Aster, and 7 days on Pacifica and Lighter. The top-5 lists and the position reads follow that order. | Aster 30d; Pacifica 7d; Lighter 7d • | Shows | Every page load | Starting value |
| Lighter and Pacifica boards: top 300Every hour, the top 300 accounts by 7-day P/L are stored from Lighter's and Pacifica's boards. Pacifica's whole board is read and re-sorted by 7-day P/L before the cut.Accounts in privacy mode take a place in the 300 but are never shown. | 300 | Shows | Hourly | Starting value |
| Only the latest board, never private accountsPacifica and Lighter lists show only the accounts on the venue's latest hourly board read, and never an account in privacy mode.An account that drops off the board leaves the lists at the next hourly read. | — | Shows | Every page load | Starting value |
| Aster trader list saved by hand Runs by hand onlyAster's trader list comes from Aster's 30-day P/L board, 100 accounts per save, and is saved only when someone runs it by hand. Every account ever saved stays on the list.Accounts that have since left Aster's board stay listed, with the rank they had when last saved. | 100 • | Shows | By hand | Fixed |
| Aster P/L windows read once a dayEach Aster trader's 24-hour, 7-, 14- and 30-day P/L is read from Aster's board once per UTC day. The lists show the latest 30-day and 7-day readings.Until a trader has a 30-day reading, the figure from the saved board is shown. | once per UTC day | Schedule | Every 2 h, each trader once per UTC day | Starting value |
| Aster positions from the latest snapshotAster positions and account value come from each trader's latest explorer snapshot, taken every 2 hours. Positions are valued at Aster's live mark, or failing that the stored mark or last price.There is no age limit: if a trader's snapshots stop, its last positions keep showing. | — | Shows | Every 2 h | Starting value |
| Aster accounts read every 2 hoursEvery 2 hours, each Aster trader on the list has its deposits, withdrawals, balances and account value read from the chain and Aster's explorer.An account in privacy mode shows no balances rather than old ones. | 2 h | Schedule | Every 2 h | Fixed |
| No liquidation price on AsterAster positions never show a liquidation price or whether they are cross or isolated margin, because Aster's explorer publishes neither.The leverage shown for Aster is the account's configured maximum for that coin, not what the position uses. | — | Shows | Every page load | Fixed |
| Aster stablecoins are collateralAster stablecoin balances are the margin behind the account's perps. They are shown, labelled as collateral, in the Aster view and left out of the all-venues holdings. | — | Shows | Every page load | Starting value |
| Aster orders matched to trades hourly Not enforcedEvery hour, stored Aster orders are matched to their fills on Aster's public trade tape, with at most 600 calls a run. Trader pages rebuild round trips from these fills.Stalled at present: the run stops with an error before matching anything whenever a market-making account is on the Aster list. | 600 calls | Schedule | Hourly | Starting value |
| Aster market data hourlyAster's mark prices, funding rates and market-wide open interest are refreshed every hour.Open interest covers the 60 busiest markets; funding intervals are measured for those 60 and assumed to be 8 hours elsewhere. | 1 h | Schedule | Hourly | Starting value |
| Positions read for each board's top 40Every hour, open positions are read for the top 40 accounts on Pacifica's and Lighter's boards. Accounts further down are not read and show 'positions not read' rather than flat.An account in privacy mode inside the top 40 is skipped, not replaced by the 41st. | 40 • | Shows | Hourly | Starting value |
| Positions from the latest sweep onlyPacifica and Lighter positions are shown only if they were read within 3 hours of that venue's latest position sweep. Each hourly sweep also removes the positions it no longer sees.The 3 hours run from the latest sweep, not from now, so if sweeps stop the last positions stay on show. | 3 h | Shows | Every page load | Starting value |
| Venue down: positions keptIf every position lookup on a venue fails in an hourly sweep, the venue is treated as down. Its stored positions are kept rather than wiped as if everyone were flat.If only some lookups fail, those accounts' positions are removed until their next successful read. | — | Shows | Hourly | Starting value |
| Positions valued at the venue's own markAster, Pacifica and Lighter positions are valued at that venue's own mark price, fetched at most once a minute. If the feed does not answer, Aster uses its hourly stored mark, Lighter keeps the value from its last hourly read, and Pacifica is valued at entry with no unrealised P/L. | 60 s • | Shows | Every page load | Starting value |
| Pacifica extra reads: 130 an hourEach hour, at most 130 extra Pacifica calls go to reading leverage settings and what open positions have realised since they opened. Leverage is read first, then positions with no figure yet, then the stalest. | 130 • | Schedule | Hourly | Owner's decision |
| Lighter P/L shown while under 2 days oldLighter's 30-day, all-time, perp and spot P/L are shown only while the reading behind them is no more than 2 days old. After that they are left blank.Without a fresh reading, a pool operator's 7-day figure falls back to Lighter's board, which includes the pool, and is flagged. | 2 days • | Shows | Every page load | Starting value |
| Lighter P/L re-read every 5 hoursA trader new to Lighter's board has its P/L read within the hour. Everyone else is re-read once their last reading is 5 hours old, oldest first, with at most 900 calls a run. | 5 h | Schedule | Hourly | Starting value |
| Lighter 30-day P/L: board figure or computedLighter's 30-day P/L is Lighter's own board figure when the trader sits in the top or bottom 2,000 of the 30-day board and runs no public pool. Otherwise it is computed from daily closes and marked as computed.A computed figure runs from the last daily close before the window opened, so it can span a little over 30 days. | 2,000 rows | Measures | Hourly | Starting value |
| Lighter perp 30-day figureLighter's 'perp 30d' is perp trading P/L, including fees and funding, over the 30 days to the last 00:00 UTC daily close. An account opened inside the window counts from zero.Spot, pool and staking P/L are left out. | 30 days to last 00:00 UTC | Measures | Hourly | Starting value |
| Lighter pools left out of a trader's bookOnly a Lighter address's own accounts, its main account and sub-accounts, count toward its positions, holdings and P/L. Public pools it runs, which hold other depositors' money, are left out, even though Lighter's boards include them.Rank on the board, and the 7-day figure when no fresh reading exists, still include the pool. | account types 0 (master), 1 (sub-account) • | Measures | Hourly | Starting value |
| Lighter margin is not a holdingLighter USDC held as perp margin is not counted as a spot holding. Only the spot wallet's balances are. | — | Measures | Hourly | Starting value |
| Lighter trade logs: favourites plus top 60Lighter trade logs are stored for starred accounts plus the 60 best-ranked Lighter accounts on record, with at most 500 calls an hour. They date open positions and list round trips on trader pages.An account that has left the board keeps its last rank here, so some of the 60 may no longer be on it. | top 60 + favourites | Schedule | Hourly | Starting value |
| Flat vs not readIn the tracker, a venue account with no open positions reads 'flat' only if its positions were read: on Aster, if it has ever been snapshotted; on Pacifica or Lighter, if it ranks in the board's top 40. Otherwise it reads 'positions not read'.A top-40 account whose read failed that hour also reads 'flat' until its next read. | — | Labels | Every page load | Starting value |
| Active filter for venue accountsFor venue accounts, the Active filter uses Aster's last stored order time. Pacifica and Lighter only reveal whether an account traded in the last 7 days, so they pass a 7-, 30- or 90-day filter only if it did, and never pass the 24-hour filter.An Aster account with no stored orders never passes the Active filter. | 7 days | Shows | Every page load | Starting value |
| Screen filters hide venue accountsFilters only a Hyperliquid wallet can meet hide every venue account, because other venues are never screened. These are a reject reason, 'screened out', or a trader type other than perp trader.The default 'kept by the screen' view still lists venue accounts. | — | Shows | Every page load | Starting value |
| Aster market makers get no trade historyAn Aster account placing 50 or more resting limit orders a day counts as a market maker if at least half of them are quoted on both sides within 60 seconds, or if it cancels at least half as many orders as it places. Its trade history is then not rebuilt, and its page says why. | 50 • | Shows | When a profile page is built (every 2 h) | Starting value first guess |
| Aster trips listed only when provenAn Aster round trip is listed only if every fill in it is sized and priced from Aster's public trade tape and nothing the explorer reports contradicts it. A market order's fill must be a run of prints within 250 ms of the order, and a limit order counts as filled once the market trades through its price. P/L is before fees and funding.Trips that closed before the first explorer snapshot are marked 'rebuilt', because nothing Aster reported can confirm them. | 250 • | Shows | At profile build (every 2 h), from fills the hourly matcher has proven; that matcher is stalled at present, so newly closed trips are not being proven | Starting value |
| Aster equity before the first snapshotBefore an Aster trader's first observed account value, the equity curve is stepped back to the starts of Aster's 7-, 14- and 30-day windows, using Aster's P/L for each window and on-chain deposits and withdrawals. A stepped-back point is kept only if it is above zero and the trading fees Aster's P/L leaves out (at most 0.04% of volume) come to no more than 10% of it.Nothing is stepped back until the trader's deposits and withdrawals have been fully scanned. | 10% • | Shows | When a profile page is built (every 2 h) | Starting value first guess |
| Aster per-coin history: 12 coinsAn Aster trader page asks Aster's board about at most 12 coins for its per-coin history: coins held now first, then the coins it ordered most in the last 30 days.Anything beyond those 12 is grouped as 'Other coins', so the table still adds up to Aster's total. | 12 • | Shows | When a profile page is built (every 2 h) | Starting value |
| Pacifica history: 120,000 fillsA Pacifica trader page reads at most 30 pages of 4,000 fills, which is the 120,000 most recent. Anything older is not rebuilt.The page says when a history was cut short. | 30 • | Shows | When a profile page is built (every 2 h) | Starting value |
| Pacifica realised-since-open: 4 pages backTo find what a Pacifica position has realised since it opened, at most 4 pages (16,000 fills on that coin) are walked back each hour. If the open is not reached, the realised figure is left blank; the open date is still Pacifica's own.A position with more than 16,000 fills on its coin since opening never gets a realised figure, because each hour's walk starts again from the newest fill. | 4 • | Shows | Hourly, within the 130-call Pacifica extra budget | Starting value |
| Pacifica page lists 120 tripsA Pacifica trader page lists at most the 120 most recent closed round trips.Summary figures such as win rate still count every round trip read. | 120 • | Shows | At page build | Starting value |