Verified record
Every settled prediction, wins and losses. Each prediction is sealed before kick-off in a table the database cannot update or delete; results are graded from the official score, and a correction is logged before it is applied.
What these numbers describe
- period
- 2026-08-29 → 2026-10-13
- cohort
- live_record_v1 · cohort_v1
- unique fixtures
- 312
- predictions / signals
- 759 / not counted
- priced / unpriced
- 525 / 234
- as of
- 2026-10-14 15:07Z
- ledger watermark
- 2026-10-14 15:03Z
- metric / record version
- perf_metrics_v1 · public_record_v1
Save this string. Each day's seal folds in the previous day's, so if any earlier day is ever rewritten, this no longer matches. The seal is our own server clock — it shows the record is unchanged, not when it was published.
Settled between 2026-08-29 and 2026-10-13
Every prediction is sealed with a hash before kick-off, in a table the database refuses to update or delete. Results are graded from the official score; a correction never overwrites silently — the previous state is written to an append-only correction log first, and the record shows it. The sealing time is WinPIQ's own server clock: no outside party has timestamped it.
Probability bucket performance
Do predictions rated X% actually win X% of the time? Computed from the settled rows below.
When we say 60%, does it happen 60% of the time?
By confidence
| confidence | n | hit rate |
|---|---|---|
| <50 | 1 | 100.0% |
| 50-69 | 62 | 48.4% |
| 70-84 | 213 | 61.0% |
| 85+ | 483 | 58.2% |
Show me the receipts 0 shown
| published | match | market | pick | prob. | conf. | price at lock | model | result | proof |
|---|