Negative Binomial Match Predictor
Model match outcomes with overdispersed scoring — better for blowout-prone matchups in baseball, hockey, and soccer
Read the guide →| Line | Over % | Over Odds | Under % | Under Odds |
|---|---|---|---|---|
| 6.5 | 66.0% | -194 | 33.5% | +198 |
| 7.5 | 56.1% | -128 | 43.4% | +130 |
| 8.5 | 46.4% | +116 | 53.1% | -113 |
| 9.5 | 37.4% | +167 | 62.1% | -164 |
| 10.5 | 29.5% | +239 | 70.0% | -234 |
| 11.5 | 22.7% | +340 | 76.8% | -331 |
| Spread | Home Covers | Home Odds | Away Covers | Away Odds |
|---|---|---|---|---|
| -0.5 | 57.6% | -136 | 41.9% | +139 |
| -1.5 | 67.5% | -207 | 32.1% | +212 |
| -2.5 | 76.0% | -316 | 23.6% | +324 |
| +0.5 | 47.2% | +112 | 52.4% | -110 |
| +1.5 | 37.0% | +170 | 62.5% | -167 |
| +2.5 | 28.0% | +258 | 71.6% | -252 |
The Negative Binomial distribution extends the Poisson model by adding a dispersion parameter (r) that controls scoring variance. While Poisson assumes variance equals the mean, real sports often have higher variance due to blowouts, rally innings, and power plays.
The r parameter: Lower r means more variance (more blowout potential). As r increases, the model converges toward Poisson. Baseball (big innings, bullpen collapses) typically has lower r than soccer.
When to use NB over Poisson: If you believe a matchup has unusual variance — e.g., a volatile bullpen in baseball, an aggressive team in hockey, or any game where blowout risk is elevated — the Negative Binomial model will produce wider scoreline distributions and different market probabilities than Poisson.
Limitations: Like Poisson, this model assumes independence between the two teams' scoring. It also uses the same dispersion for both teams — in practice you could set different r values per team, but this simplified version uses a shared r for clarity.