What if? · Home advantage
Does home advantage exist in every sport?
The home side wins more often than the away side in every head-to-head sport, and by a similar margin.
Provisional: with 3 sports the spread between them is only loosely estimated, so a high score means the sports are consistent with one shared effect, not proven identical.
Home team's share of games that had a winner
| Sport | Games | Home wins | Away wins | Draws | Home share | 95% CI |
|---|---|---|---|---|---|---|
| American Football | 544 | 291 | 252 | 1 | 54% | 49–58% |
| Basketball | 3849 | 2100 | 1749 | 0 | 55% | 53–56% |
| Football | 381 | 163 | 114 | 104 | 59% | 53–64% |
Does it help predict games?
A one-number model: learn the home share from the first 70% of each sport's games, then predict the rest. Lift is the improvement in log loss over a coin flip.
| Sport | Test games | Coin flip | Home model | Lift |
|---|---|---|---|---|
| American Football | 163 | 0.693 | 0.694 | -0.2% |
| Basketball | 1155 | 0.693 | 0.689 | 0.6% |
| Football | 84 | 0.693 | 0.675 | 2.6% |
The conventional wisdom
Home advantage is treated as a fact of sport - crowd, travel and familiarity all favour the host.
How we calculated this
Regular-season games only. For each sport we count home wins and away wins, set draws aside, and express the home side's edge as a log odds with a 95% confidence interval. The sports are then pooled with a DerSimonian-Laird random-effects model. The universality score is the share of the effect that is common to every sport: pooled effect squared, divided by itself plus the between-sport variance. Neutral-site regular-season games (NFL international games, for example) cannot yet be told apart from true home games and are counted as home games.
Last computed 12 November 2026.