Edge radar
The public record cut by segment: every settled prediction with a logged closing line, graded model-vs-market by pool, favourite band, competition and market. A segment counts as a validated edge only when a real out-of-time sample beats the de-vigged closing line and a paired bootstrap says the gap is not noise. Everything else is reported as it is — not an edge.
How a segment earns the label
Out-of-time sample: only predictions stamped before kickoff and settled afterwards count — never the data the model was trained on. Each prediction's RPS is paired with the RPS of the closing line on the same match, so the two are scored on exactly the same rows.
Paired bootstrap: a segment needs at least 30 graded matches; its mean gap (model RPS minus closing-line RPS) must be below −0.001, and the upper bound of a one-sided 95% stationary-bootstrap interval on that gap — 2,000 resamples that respect the matchweek ordering — must stay below zero. Below the sample floor the verdict is 'not validated', which is neither a win nor a loss.
A validated edge is a statement about the past sample it was measured on. It is not a guarantee, and a segment can lose the label when new matches settle.
Validated edges
The segments that currently clear the gate, best first. If this list is empty, that is the honest state of the record — the tables below still show every segment and why it did not qualify.
No segment currently clears the bar. That is the state of the record, not a gap in the page.
1X2 by segment
All 1X2 calls, then split by pool (club or national-team ratings, kept apart by design) and by how strong the model's favourite was.
| Segment | Settled | Graded | Model RPS | Closing RPS | Gap | Verdict |
|---|---|---|---|---|---|---|
| All 1X2 calls | 4 | 0 | — | — | — | Not validated — no closing line stored |
| Clubpool | 4 | 0 | — | — | — | Not validated — no closing line stored |
| Leanfavourite band | 3 | 0 | — | — | — | Not validated — no closing line stored |
| Toss-upfavourite band | 1 | 0 | — | — | — | Not validated — no closing line stored |
By competition
| Segment | Settled | Graded | Model RPS | Closing RPS | Gap | Verdict |
|---|---|---|---|---|---|---|
| Ligue 1 | 1 | 0 | — | — | — | Not validated — no closing line stored |
| Primeira Liga | 1 | 0 | — | — | — | Not validated — no closing line stored |
| Serie A | 1 | 0 | — | — | — | Not validated — no closing line stored |
| La Liga | 1 | 0 | — | — | — | Not validated — no closing line stored |
By market
Each market is scored against its own stored closing line. RPS for a two-outcome market (BTTS, over/under) sits on a different scale from three-outcome 1X2, so rows here are never compared with each other — only with their own line.
| Segment | Settled | Graded | Model RPS | Closing RPS | Gap | Verdict |
|---|---|---|---|---|---|---|
| 1X2 | 4 | 0 | — | — | — | Not validated — no closing line stored |