Why are so many addresses rejected?
All-time PnL distribution of the census (raw leaderboard, no gates)
| bucket | wallets |
|---|---|
| pnl > 0 | 22,632 |
| pnl >= 50k | 14,315 |
| pnl >= 100k | 10,732 |
| pnl >= 250k | 6,621 |
| pnl >= 500k (gate) | 4,325 |
| pnl >= 1M | 2,630 |
| pnl >= 5M | 698 |
Sequential funnel, recounted from the raw file
| gate | in | cut | cut share | out |
|---|---|---|---|---|
| alltime_pnl | 45,107 | 40,782 | 90.4% | 4,325 |
| alltime_roi | 4,325 | 638 | 14.8% | 3,687 |
| account_value | 3,687 | 2,311 | 62.7% | 1,376 |
| month_pnl_not_positive | 1,376 | 86 | 6.2% | 1,290 |
| week_pnl_not_positive | 1,290 | 1,033 | 80.1% | 257 |
| pnl_per_vlm_too_low | 257 | 6 | 2.3% | 251 |
Each gate alone, on the 4,325 wallets with all-time PnL ≥ $500k
| gate | would remove |
|---|---|
| roi < 0.5 | 638 |
| account value < 1M | 2,580 |
| month pnl <= 0 | 600 |
| week pnl <= 0 | 3,303 |
| pnl/vlm < 5bps | 84 |
Sensitivity: pass-1 survivors under alternative rules
| rules | survivors |
|---|---|
| current rules | 251 |
| drop the week-pnl gate | 1,279 |
| account value as a label (no 1M gate) | 754 |
| both of the above | 3,169 |
| pnl >= 250k instead of 500k | 255 |
| pnl >= 250k + drop week gate + no 1M gate | 4,700 |
| pnl >= 100k + drop week gate + no 1M gate | 7,291 |
Pass 2 (portfolio proxies)
| item | value |
|---|---|
| pass-1 survivors checked | 241 |
| best day > 40% of month pnl (gate) | 115 |
| best day > 60% of month pnl | 73 |
| best day > 100% (rest of month net negative) | 28 |
| rejected by reason | {"perp_month_pnl_not_positive": 76, "positive_day_share": 17, "max_day_share": 83, "perp_week_pnl_not_positive": 16, "staker": 1, "labeled_hub": 1} |
Regenerate with python3 funnel_breakdown.py --json reports/funnel_breakdown.json.