—
| Variant | Win rate | Days | Avg day trades | Avg day | Median day | Avg win | Avg loss | Best day | Worst day | CVaR 5% | Out of sample | Total |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Loading… | ||||||||||||
Out of sample is the held-out second half of the history (split —) — the only performance number here that was not used to choose the variant, and the column the table is ranked on by default. n is how many held-out sessions it rests on; marks a sample too thin to read as evidence. Hover it to see the fit half and how far the result decayed; marks a material drop and in-sample only marks a variant that was positive on its own selection data and negative once held out. The count above the table is the variants positive in both halves, which is a stricter and more useful bar than positive out of sample alone. CVaR 5% is the average of the worst 5% of days — what a bad session costs, which for short-premium trades matters more than the average.