portfolio policy replay (predictions of backtest #16)
← lab · n bets 5,959 · hit rate 0.3536 · source run 16 · ran 7/4/2026
What this is: A one-off study — its numbers below are exactly what the run reported.
| note | same bucket/blend/quarter-Kelly policy as the live portfolio_v1; opening prices; walk-forward probs |
| n bets | 5,959 |
| hit rate | 0.3536 |
| source run | 16 |
| max drawdown | 0.9852 |
| total staked | 102007.04 |
| final bankroll | 268.9 |
| start bankroll | 1,000 |
| roi on turnover | -0.0072 |
by bucket
| value | {"n":3217,"pnl":397.64} |
| ballast | {"n":1619,"pnl":-849.97} |
| longshot | {"n":1123,"pnl":-278.76} |
Reading the columnswhat each number actually means
| AUC | predictability: 0.50 = coin flip, ~0.70 = ceiling for sports |
| Importance | how much the model leans on this factor (permutation importance) |
| Direction | sign of the raw correlation with the outcome |
| Survives all eras | effect points the same way in every historical era |
Spec · the reproducible recipe
{
"kind": "portfolio_backtest",
"source_run": 16,
"start_bankroll": 1000
}