Football, basketball and baseball are calculated differently
Scoring works differently in each sport, so the maths does too. What approach fits which sport, and why.
· 5 min read
"Predicting a match" sounds like one problem. It is three, because goals, points and runs do not behave the same way.
Football — low scoring, so the distribution matters
A typical match has two or three goals. With numbers that small, a single goal swings the result, and randomness is large relative to the total.
The standard approach treats goals as a Poisson process — events arriving at some rate, independently. You estimate each side's attacking and defensive rate, adjust for home advantage, and get a distribution over scorelines rather than a single prediction.
The known weakness: goals are not fully independent. A team that goes 2–0 up changes how it plays. Dixon-Coles is the common correction — it adjusts the low-scoring results (0–0, 1–0, 1–1) where the independence assumption breaks worst.
Why this matters to a reader: in football, a good model does not tell you who wins. It tells you how often each scoreline happens.
Basketball — high scoring, so it behaves like a normal curve
A hundred possessions a side, two hundred points in the game. With numbers that large, the noise averages out and the distribution becomes roughly normal.
That changes the toolkit entirely. Instead of modelling events, you model efficiency per possession and pace:
- Points scored per 100 possessions
- Points allowed per 100 possessions
- How many possessions the teams generate
Pace is the underrated one. Two efficient teams playing slowly produce a low total. Two mediocre teams playing fast produce a high one. Efficiency predicts who wins; pace predicts how much is scored.
Baseball — a sequence of isolated duels
Baseball is not continuous. It is a chain of discrete matchups: this pitcher against this batter, with a known base state and a known out count.
That structure is a gift. You can model each plate appearance separately and chain them, which is why baseball analytics is decades ahead of the other two.
It also means one player dominates the variance. A starting pitcher touches every plate appearance for six innings. No single footballer or basketball player has that leverage.
Practical consequence: in baseball, the starting pitcher matchup is the single largest input. In football, no individual comes close — except the goalkeeper, occasionally.
Side by side
| Football | Basketball | Baseball | |
|---|---|---|---|
| Scoring events | 2–3 | ~200 points | 4–10 runs |
| Usual model | Poisson (+ Dixon-Coles) | Efficiency × pace | Plate-appearance chains |
| Biggest single input | Team strength | Efficiency gap | Starting pitcher |
| Randomness | High | Low | Medium |
| Home advantage | Large | Moderate | Small |
What this means when you read a number
A 60% win probability means something different in each sport.
In basketball, 60% is a real edge that shows up reliably over a season, because the noise is low. In football, 60% still loses four times in ten, and a run of those four is completely ordinary.
The same number carries different confidence depending on the sport. That is the part most people skip.
BetScore applies a different model per sport rather than one formula for everything.