Rule favorite_hold at threshold 0.8, category all, window 2025-09-11 to 2026-09-10: matched 6281 of 20000 candidates, hit rate 0.8195, Brier score 0.1483. The universe was capped at 20 000 markets by volume; only the largest by volume entered the run.
| Bucket | n | Implied | Realized | Mean return |
|---|---|---|---|---|
| 0.80-0.85 | 2026 | 0.8189 | 0.7557 | -0.0772 |
| 0.85-0.90 | 1470 | 0.8694 | 0.7857 | -0.0961 |
| 0.90-0.95 | 1520 | 0.9207 | 0.8507 | -0.0761 |
| 0.95-1.00 | 1265 | 0.9654 | 0.9233 | -0.0437 |
Rule longshot_hold at threshold 0.1, category all, window 2025-09-11 to 2026-09-10: matched 4616 of 20000 candidates, hit rate 0.1218, Brier score 0.1091. The universe was capped at 20 000 markets by volume; only the largest by volume entered the run.
| Bucket | n | Implied | Realized | Mean return |
|---|---|---|---|---|
| 0.00-0.05 | 1269 | 0.0318 | 0.0701 | 1.2573 |
| 0.05-0.10 | 2457 | 0.0728 | 0.14 | 0.9832 |
| 0.10-0.15 | 890 | 0.1 | 0.1449 | 0.4494 |
How to read this: implied is the average entry price in a bucket, realized is the share of those positions that paid out, so a bucket where realized exceeds implied beat its price and one where it trails fell short of it.
Replay any rule, category and window with POST /v1/polymarket/backtest at $0.25 per call.