Modelling player fouls: every foul is two stats
By Adam · Updated 16 July 2026 · 6 min read
Fouls are the unfashionable corner of the bet builder — no highlight reel, no fantasy points, barely a mention on the podcasts — and they're where this model does some of its best work. The reason is a quirk of bookkeeping with real consequences: every foul is two stats. The tackle that goes wrong is a foul committed for one player and a foul won for the other, two markets priced separately from the same collision. This is the walk through how we model both sides — companion to the shots and cards guides, and a plain-English disclosure of how the fouls and fouls won models work.
The same spine, again
The machinery will be familiar if you've read the other modelling guides: a per-90 rate with minutes as the denominator, recent matches weighted more on a half-life, and thin samples shrunk toward a positional prior until the player has earned his own number. Fouls are a middle-frequency stat — the average player commits about one per 90 — so the rates stabilise faster than cards and slower than shots. What changes is everything around the rate.
Who actually fouls
The folklore says defenders foul. The data says the foul lives in midfield and the press: midfielders commit about 1.2 fouls per 90 and forwards are right behind them at roughly the same rate — modern forwards are the first line of the press, and pressing is how fouls happen — while defenders trail at under 0.9. The winning side has a clearer hierarchy still: forwards win the most fouls (about 1.3 per 90), because dribblers and hold-up strikers spend their evenings being kicked, and defenders win barely half that. So a thin-sample player gets priced as his position first, and the position matters more here than the market seems to think.
Committed and won are mirror images
Here's the structural elegance. When we price a player's fouls committed, the opponent adjustment isn't some abstract aggression score — it's how many fouls that opponent wins, relative to the league. A team of quick, tricky dribblers drags fouls out of whoever it plays. And when we price fouls won, the adjustment flips to how many fouls the opponent commits. The two markets are mirror images, and each one's opponent factor is the other one's stat. The spread is meaningful — roughly 0.8× to 1.2× either way — which is the difference between a coin-toss 1+ line and a strong favourite.
Why there's no referee factor
Unlike cards, the fouls model deliberately carries no referee factor. The referee's big, measurable, price-moving effect is on what a foul becomes — a talking-to or a booking — and that lever lives in the cards model, where it's the headline. Folding a noisy referee estimate into fouls as well would mostly add variance. It's the same discipline as leaving derby heat out of shots: a factor has to earn its place with evidence, not vibes.
The distribution does the rest
Rate × minutes × opponent gives an expected foul count; the step to a probability runs through the Negative Binomial, because real foul counts are lumpier than a well-behaved Poisson — a player can spend one match chasing shadows and the next in an armchair. Same logic as the shots guide: get the spread right or the 2+ lines, where the interesting prices live, will quietly rob you.
What the record says
On the public track record, the 1+ fouls markets are among the model's strongest: fouls committed beats the baseline by around 14% and fouls won by around 15%, each across 9,521 out-of-sample predictions. The honest explanation isn't that fouls are easy — it's that they're ignored. Attention and sharp money concentrate on goals and shots; foul lines are priced coarsely by comparison, and a model that simply does the arithmetic carefully — the right denominator, the right prior, the right opponent — is doing more than the market bothered to do. That's the whole edge, and it lasts only as long as the market keeps ignoring these lines.
Frequently asked
What counts as a foul in these markets?
A foul the player concedes (fouls committed) or is awarded (fouls won), per the match-data provider. They're separate markets settled separately, and definitions can differ slightly between providers and bookmakers — worth one read of the rules before it matters.
Why do forwards commit so many fouls?
The press. Modern forwards are the first line of defence, and pressing is how fouls happen — forwards commit fouls at nearly a midfielder's rate, roughly 40% above defenders. The folklore that fouling belongs to centre-backs is a generation out of date.
Does the referee affect foul counts?
The model deliberately leaves him out here. His big, measurable, price-moving effect is on what a foul becomes — a booking — and that lever lives in the cards model, where the evidence is strongest. Folding a noisy referee estimate into fouls as well would mostly add variance.
Why are fouls good value markets?
Because they're ignored. Attention and sharp money concentrate on goals and shots, so foul lines are priced coarsely by comparison. On the public track record the 1+ fouls markets are consistently among the model's strongest results.
Keep going
SharpXI models probabilities; it doesn't promise profit, and neither should anyone else. Betting involves risk — never stake more than you can afford to lose. 18+ — please gamble responsibly.