Track record
Every prediction, logged before kick-off.
The methodology page reports a historical backtest. This page is different: each row below was written to a public log before its match was played, then graded against the real result once it existed. Nothing here can be tuned in hindsight.
A small sample proves little either way. Read the accuracy figures as a running check on the
model's honesty, not as evidence of an edge — that question is only answered by
closing line value over hundreds of graded predictions.
Model vs market, settled predictions only
3
Graded predictions
0.2095
Model RPS
100.0%
Model accuracy
0.1963
Market RPS (n=3)
100.0%
Market accuracy
120
Open (awaiting kick-off)
Lower RPS is better. Bookmaker closing odds are the benchmark that matters — see how it's measured.
By league
| League | n | Model RPS | Accuracy | Market RPS |
|---|---|---|---|---|
| E0 | 3 | 0.2095 | 100.0% | 0.1963 |
Recently settled
| Kick-off | Home | Away | Pick | Result | RPS | |
|---|---|---|---|---|---|---|
| 14 Sep 2026 |  Leeds |
 Newcastle |
H | 4–1 (H) | 0.235 | ✓ |
| 13 Sep 2026 |  Man United |
 Man City |
A | 0–1 (A) | 0.229 | ✓ |
| 13 Sep 2026 |  Coventry |
 Brighton |
A | 0–5 (A) | 0.164 | ✓ |
Machine-readable: /api/track-record.