Macfax Adjusted Ratings
How Macfax estimates team strength
Adjusted Ratings are the backbone of Macfax team strength estimates. They measure how efficiently a team scores and prevents scoring per possession, then correct those raw numbers for opponent quality, game location, and sample size. The result is a fairer, more stable basis for comparing teams across wildly unequal schedules.
What It Measures
Four values, all expressed per 100 possessions on a neutral-site basis against an average Division I opponent. AdjO (Adjusted Offensive Efficiency) measures points scored. AdjD (Adjusted Defensive Efficiency) measures points allowed — lower is better. AdjEM (Adjusted Efficiency Margin) is AdjO minus AdjD, the single best summary of overall team quality. AdjTempo is projected possessions per 40 minutes.
Why It Matters
Points per game lies. A team scoring 80 against a weak non-conference schedule is not the same as one scoring 80 against elite opponents every night. Raw win-loss records suffer from the same distortion. Adjusted Ratings remove schedule context, letting you ask a more honest question: how good is this team on a neutral floor against equal competition?
How to Interpret
AdjEM is the number to watch. Positive means the team outscores opponents per possession; negative means they get outscored. National average is 0 by definition — it is a margin above or below the mean. AdjO and AdjD are centered around the national scoring average — roughly 100 points per 100 possessions in a typical season. An AdjO of 118 is elite offense. An AdjD of 94 is elite defense.
Formula
Raw OE = 100 × Points / Possessions Raw DE = 100 × Opp Points / Opp Possessions AdjEM = AdjO − AdjD
Technical Notes
- Opponent adjustment is iterative — each team's rating depends on opponents' ratings, which depend on their opponents' ratings, and so on. The process runs until ratings converge.
- Location normalization adjusts home and away games to a neutral-site baseline using site factors derived from league-wide patterns. The exact calibration is internal to Macfax.
- Bayesian shrinkage pulls early-season ratings toward the national average. The pull weakens as games accumulate — ratings stabilize meaningfully after around 15–20 games.
- More recent games carry somewhat more weight to reflect current form. The model does not discard earlier results, but recency is factored in throughout the season.
- Extreme mismatches receive less weight than close, competitive games. Games that are unexpectedly close relative to the expected gap receive modestly more weight.
- Only games against Division I opponents count toward ratings.
- Exact calibration values for shrinkage, recency decay, and site adjustment are internal and may be recalibrated as validation data accumulates.
- •Early-season ratings (first 10–15 games) carry high uncertainty. Treat them as directional, not definitive.
- •Injuries, suspensions, and lineup changes are not automatically reflected — the model only knows what happened on the court.
- •Non-conference schedule imbalance persists in adjusted ratings when a team has very few crossover games against diverse opponents.
- •Possession estimates from box-score data are approximations. Play-by-play counts are more precise but not always available.
- •Public box-score data occasionally contains errors. Corrections are applied on the next update cycle.
A team with AdjO 118.5 and AdjD 94.2 has AdjEM +24.3 — they outscore opponents by 24.3 points per 100 possessions on a neutral floor against average Division I competition. In a typical season, this places them among the top five or ten teams nationally. A team with AdjO 105.0 and AdjD 104.8 has AdjEM +0.2 — barely above average, outscoring opponents by less than one possession per 100. These numbers live on the same scale, making cross-conference comparison direct.