Direct Answer
Regression in statistics describes how a dependent variable changes with one or more predictors. In gambling models it is the workhorse method for translating observable data (offense rating, weather, rest) into a probability or point projection.
Key Takeaways
- Linear regression for continuous outcomes.
- Logistic regression for win/loss.
- Regression to the mean punishes naive projection.
Where regression fits
Linear regression for continuous outcomes (point totals). Logistic regression for binary outcomes (win/loss). Both are interpretable, fast, and surprisingly hard to beat with more complex methods in many sports settings.
Regression to the mean
A separate but related concept: extreme observations tend to be followed by less extreme ones. Hot streaks regress. Cold streaks regress. Sports projection systems that ignore this consistently overshoot.
Frequently asked questions
Do I need machine learning to model sports?+
No. Well-built regression models beat poorly built ML models in nearly every sports application.
Educational only. Not wagering, financial, or legal advice. See our editorial policy.