Macfax Matchup Model
Game projections, spreads, totals, and win probability
The Macfax Matchup Model projects game outcomes by combining each team's opponent-adjusted offensive and defensive efficiency, estimated game pace, and site factors into a projected score, spread, total, and win probability. It also computes per-dimension Four Factor edges for the specific matchup. The model is built on multiplicative efficiency interaction — how one team's offense interacts with the opponent's defense, normalized to the national average.
What It Measures
The Matchup Model estimates the most likely score and outcome for a specific game. It produces a projected score for each team, a projected spread, a projected total, and a win probability. Separately, it computes Four Factor matchup edges — showing which team has the structural efficiency advantage on shooting, turnovers, rebounding, and free throw rate for that specific pairing.
Why It Matters
Game previews without quantitative projection rely on narrative. The Matchup Model converts team ratings into specific, falsifiable predictions that can be evaluated after the game. Because the projection is built on opponent-adjusted inputs, it correctly accounts for schedule strength — a high-scoring team facing an elite defense is projected differently than when they faced an average defense.
How to Interpret
Projected spread is the model's best estimate of the margin on a neutral or home-site basis. Win probability is expressed as a percentage — 70% means the model expects the favored team to win roughly 7 out of 10 times under similar conditions. A 70% win probability is not a lock; basketball variance is high enough that the underdog wins 3 in 10. Do not treat any probability under 90% as certain. Four Factor edges show which team has the advantage on each possession dimension — useful for understanding why the model favors one team.
Formula
Proj OE_A = (AdjO_A × AdjD_B) / NatAvg Proj OE_B = (AdjO_B × AdjD_A) / NatAvg Site factors applied symmetrically to home and away teams. Expected pace = weighted blend of both teams' adjusted tempos. Expected Score = Proj OE × (Expected Possessions / 100) Projected Spread = Expected Score_A − Expected Score_B
Technical Notes
- Efficiency interaction is multiplicative: projected offense is the product of the offensive team's AdjO and the defensive team's AdjD, divided by the national average. This naturally scales with both teams' quality rather than treating efficiency as additive.
- Game pace is estimated as a weighted blend of both teams' adjusted tempos, producing a game-specific possession estimate rather than using either team's tempo alone.
- Home court factors are applied symmetrically — a split of the total advantage is applied to the home team's offense and removed from the away team's offense. The exact calibration is internal to Macfax.
- Win probability is derived from the projected spread using a Normal CDF with a calibrated standard deviation that reflects typical game-to-game variance in Division I basketball.
- Four Factor matchup edges are also computed multiplicatively per dimension — projecting the eFG%, turnover, rebounding, and FTR margins expected for this specific game given both teams' adjusted factor profiles.
- A volatility score is computed separately for each matchup to indicate how much more uncertain this game is relative to the average projection. Volatility reflects factors like game pace, three-point volume, and recent performance consistency.
- Recent form trends are displayed in the matchup context as supplemental information. They are not inputs to the projected spread or win probability calculation.
- •Recent form does not affect the projected spread. The model's projection is based on season-to-date adjusted ratings. Recent form is shown alongside but does not shift the number.
- •The model does not adjust for injuries, suspensions, or roster availability unless ratings already reflect those absences from prior games.
- •Travel, rest, altitude, and other logistical factors are not currently modeled.
- •Late lineup changes or game-time decisions before tip-off are not reflected.
- •Early-season projections carry more uncertainty because adjusted ratings have higher variance with fewer games played.
- •Win probability assumes a calibrated variance around the projection. Neutral-site tournament games may have different variance characteristics than regular-season games.
Illustrative: Team A (AdjO 116.2, AdjD 95.1) hosts Team B (AdjO 108.4, AdjD 99.7), national average 100.0. Proj OE_A = (116.2 × 99.7) / 100.0 ≈ 115.8. Proj OE_B = (108.4 × 95.1) / 100.0 ≈ 103.1. With expected 68 possessions and home court applied, model projects Team A 78 – Team B 70, spread −8, total 148, win probability 77% for Team A.