NBA Win Predictor

How the NBA Win Predictor Works

The predictor uses Bill James' Log5 formula, first developed for baseball and widely adopted in basketball analytics. The formula calculates the true win probability between two teams by considering their win percentages relative to a baseline .500 team. Unlike a simple difference in records, Log5 accounts for the non-linear relationship between team quality and win probability.

Formula: P(A beats B) = (A − A×B) ÷ (A + B − 2×A×B) where A and B are decimal win percentages. A .650 team playing a .450 team produces a 68.4% win probability for the better team — notably stronger than just comparing the raw 20-point record gap.

Home Court Advantage

When home court is selected, the home team's win percentage is adjusted by +5 percentage points before Log5 is applied — reflecting the NBA's historically observed home win rate of approximately 59–60%. This shifts the probability meaningfully in close matchups and accounts for factors like crowd energy, travel fatigue, and familiarity with the home court.

7-Game Series Probability

Given a single-game win probability, the model calculates the probability of winning a best-of-seven playoff series using the binomial formula. The series can end in 4, 5, 6, or 7 games. Notably, even a team with a 60% single-game advantage only wins the series about 71% of the time — illustrating how NBA playoffs compress team quality differences over a small sample. A true coin-flip team (.500) always wins exactly 50% of series regardless of length.

Point Spread Estimate

The estimated spread uses a linear approximation: each 3% shift in win probability translates to roughly 1 point of spread advantage. This is a simplified model — actual betting lines incorporate sharper line movement, public money, and real-time injury information. The estimated spread is an analytical reference point, not a betting recommendation.