Net flow comes from the exchange’s own accounting (F = ΔaccountValue − ΔpnlHistory), not the ledger — the ledger misses perp↔spot moves on dex-abstraction accounts entirely. 73 equity points on file.

The P&L line is mark-to-market — it comes from the exchange’s own pnlHistory and already carries open-position profit, not just closed trades. What it is not is continuous: portfolio returns about a hundred points whatever the account’s age, so this wallet is sampled every 12.3 days and any drawdown that opened and closed between two samples is invisible. Every drawdown figure above is a floor, not the true worst. Continuous intraday drawdown only starts accumulating from the position tracker, which has been running since 12 Sep. The two percentages divide the same dollar loss by different things: There is no single right denominator, so both are named rather than one being picked: the largest the account was ever seen to hold, and every dollar ever committed to it. Either can pass 100% on a leveraged account — this wallet gave back $21.13M against a peak P&L of $88.34M. Dividing by peak P&L is deliberately not shown as a headline: on a wallet that was barely profitable before losing, it produces numbers in the thousands of percent. A drawdown measured on raw account value is deliberately not shown at all: this wallet withdrew $49.44M, so its account value fell −61.9% from peak — a figure that reads like a blow-up and is nothing of the kind. Every drawdown above is measured on P&L, which withdrawals cannot move.

A round trip is flat→flat: fill → order → decision → campaign, with same-millisecond fills re-chained by startPosition and TWAP slices merged by tid. In and Out are when it opened and closed, with the size-weighted average price of the fills each way where it is known; P/L is before fees. A trip marked censored was already open where the history starts, so its entry is not known. A trip marked open is still held: its P/L is what it has made so far, the parts closed plus the live unrealised figure.