Each live trade plots your cumulative net P&L (yellow, auto-loaded). Inside green = the edge is realizing as the 6-year backtest predicts. Amber is the stress case — even if live delivers only half the backtest edge, you'd be here, and it still separates from luck by n*=80 trades (~28 days). Gray is pure luck. The honest read: a sustained slide below the amber band is the real signal the edge isn't transferring — not red days. You'll know by ~20 trades (~7 days) at full edge, ~80 even at half.
| after N trades | ~days | p5 (unlucky) | median | p95 (lucky) | half-edge median | luck p95 | P(green) |
|---|---|---|---|---|---|---|---|
| 10 | 4 | $725 | $1,673 | $3,179 | $786 | $1,405 | 100% |
| 20 | 7 | $2,004 | $3,428 | $5,506 | $1,653 | $1,957 | 100% |
| 30 | 11 | $3,386 | $5,213 | $7,665 | $2,551 | $2,342 | 100% |
| 50 | 18 | $6,295 | $8,792 | $11,852 | $4,356 | $2,979 | 100% |
| 100 | 35 | $14,037 | $17,677 | $21,936 | $8,805 | $4,192 | 100% |
| 250 | 88 | $38,368 | $44,287 | $50,809 | $22,107 | $6,448 | 100% |
Caveats: tests realization, not validity (validity already established: gold-val DSR/White's/permutation + 0 neg-months/73). Assumes regime resembles 2020-26; decay shows as sliding out the bottom (the live drift/CUSUM watches this). IID bootstrap (justified: ~2.8 trades/day, 0 neg-months = low clustering). 2/2/3 sizing (4/4/6 ≈ 2×). Built from the canonical 4,358-trade ledger; 4/4/6 gross moMed ~$20.6k → net ~$20.2k (~2% above the $19.8k net085 summary, a net-convention diff). Page auto-loads your live curve from monitoring.json; band is static.