Public Track Record
100% transparent historical accuracy computed directly from settled match results.
A win rate alone cannot tell you whether a stated 70% really means 70%. Each band below compares the confidence we published against how often those picks actually landed. Landing inside the shaded band is the target.
Gap is measured in percentage points against the mid-point of the claimed band. Bands marked thin hold fewer than 10 settled picks and are too small to read much into yet.
Only picks published before kickoff are counted. Predictions are locked at publication and cannot be edited once a fixture has started.
How the engine was tested
Before a single pick was published, the model was replayed over 34,993 finished matches from 2014–2026. Each one was predicted using only what was known the day before it kicked off — no result the model had already seen, no hindsight.
| Market | Picks tested | Landed | Tipping blind | Difference |
|---|---|---|---|---|
| Match Winner | 12,622 | 65.0% | 44.0% | +21.0 pts |
| Over 2.5 Goals | 24,262 | 57.0% | 52.1% | +4.9 pts |
| Both Teams To Score (GG) | 20,266 | 55.2% | 53.0% | +2.3 pts |
| Over 8.5 Match Corners | 24,010 | 67.0% | 64.8% | +2.2 pts |
| Over 2.5 Match Yellow Cards | 15,315 | 82.3% | 73.3% | +9.0 pts |
Read the last column. A hit rate on its own tells you almost nothing. Roughly three matches in four produce three bookings, so a card tip that lands 80% of the time has beaten nothing at all. “Tipping blind” is what you would have scored by backing the commoner side of that market every single week, and the difference is the only part the model is responsible for.
These figures cover the picks that clear each market’s publication threshold — the ones that reach this site. Below that line the engine keeps its opinion to itself.
Across the whole test the model projected 2.75 goals a match against 2.75 actually scored.
Past performance over historical fixtures is not a forecast of future results, and no model can guarantee an outcome. Football is decided by deflections, red cards and refereeing calls that no amount of data anticipates.
