2026 NBA Finals Prediction Engine
KNICKSvs.SPURS
We use three different prediction engines to simulate the 2026 NBA Finals between the New York Knicks and San Antonio Spurs. See how we created the prediction models on Medium
PACE-ADJUSTED EFFICIENCY MODEL
This is engine 1: the pace-adjusted efficiency model. It is the simplest of the three predictors. It uses only each team's offensive rating, defensive rating and pacing to predict the each team's probability of winning the 2026 NBA Finals after running 10,000 monte carlo simulations.
ELO POWER RATING MODEL
This is engine 2: the elo power rating model. It rates every team with an elo system built game by game across the 2025-26 season. It then combines that with each finals team's pythagorean expectation (an estimate of how many games a team "should win" based on their offesive and defensive ratings) to predict each team's probability of winning the series after running 10,000 monte carlo simulations. The elo system is a better metric than engine 1 because it rewards beating good teams with higher power ratings and it adjusts for margin of victory.
FOUR-FACTOR PLAYER IMPACT
This is engine 3: the four factor player-impact model. This is the most complex of the three predictors. It layers player-level projections on top of team-level four factors to produce a more granular prediction that can also be used to explore how the outcome of the series is affected when players are injured or playing time varies.
The layers are as follows:
- 1. Layer One: We calculate each team's efficiency rating using Oliver Dean's four factors: Effective Field Goal Percentage (eFG%), Turnover Percentage (TOV%), Offensive Rebounding Percentage (ORB%), Free Throw Attempt Rate (FTA Rate).
- 2. Layer Two: We score each player's individual offensive efficiency using their true shooting percentage and usage rate.
- 3. Layer Three: We adjust each player's individual offensive efficiency against the strength of their defender. You can view the matchups below.
- 4. Layer Four: We weight each player based on the number of minutes they are expected to play in the finals. You can adjust each players minutes to see how the series win probability for each team changes.
- 5. Layer Five: We combine layers 1-4 into a a single win probability for both teams using a logistic regression model. We then run 10,000 monte carlo simulations to simulate the series.