How this works
Nothing here is proprietary. The model is a few dozen numbers and they are all published on this site.
Matches trained
96,903
rolling, refreshed nightly
Operators
47
active enough to rate
Error per match
1.854
goals, on unseen data
Noise floor
1.860
best achievable, same data
The model
Each side’s scoring rate is built from who is playing:
rate(home) = base + attack(home) + defence(away) + home term + warm-up(home)
rate(away) = base + attack(away) + defence(home) + warm-up(away)
Those rates become a distribution over total goals, which is where every over/under number on this site comes from. The distribution is deliberately narrower than a textbook Poisson: goals here are less erratic than the standard assumption, so the shape is compressed by a fitted factor, currently 1.120.
Three things that surprised us
- Home advantage is nothing. The fitted home term is +0.0194 on the log scale, about two hundredths of a goal. “Home” is only a stream slot. Any model ported from real football would carry a bias fifteen times too large.
- Players warm up, they do not tire. More matches into a session means more goals, worth roughly 0.6 across a long day. Every operator has their own curve.
- The scoring level drifts more than the players matter. It swung 0.85 goals over the year — about five times the value of knowing who is on the pitch.
Is it any good?
Judge it on calibration, not on how often it picks a winner. This is every prediction on 15,486 matches the model never saw during development, grouped by how confident it was. If the two columns match, a stated probability means what it says.
| We said | It happened | Matches |
|---|---|---|
| 27% | 28% | 216 |
| 36% | 33% | 668 |
| 46% | 43% | 2,227 |
| 55% | 55% | 4,004 |
| 64% | 63% | 4,503 |
| 75% | 74% | 3,025 |
| 83% | 80% | 811 |
What it cannot do
- It is already at its limit. Average error is 1.854 goals against a floor of 1.860. The model is as accurate as its own uncertainty allows; the rest is what a 15-minute match does on its own.
- High-scoring calls run hot. When it says a match is very likely to go high, it is a little over-confident. Trust the middle of the range more than the top.
- It knows nothing about prices. No odds are used anywhere. A well-calibrated probability still loses money against a bookmaker’s margin, so nothing here says whether a price is worth taking.