🐻⬇️🏀

2025-2026 NCAAM Model Performance Analysis

Scope
All Games Ranked Games

All scored games in the selected league and season. AP Poll is excluded here.

Season
2025-2026 2024-2025 2023-2024 2022-2023 2021-2022 2020-2021 2019-2020 2018-2019 2017-2018 2016-2017 2015-2016 2014-2015 2013-2014 2012-2013 2011-2012 2010-2011 2009-2010 2008-2009 2007-2008 2006-2007 2005-2006 2004-2005 2003-2004

Comparing prediction accuracy across 2975 games using multiple rating models.

Model Catalog

7-day holdout coverage: 16/17 models .

Rolling Holdout Curves

Each point is a strict weekly holdout: train on all games before that week, test on that week. This first version uses a 21-day warmup, then 7-day holdouts stepped forward weekly.

Log Loss Brier AUC Accuracy

Weekly strict holdout log loss. Lower is better. Showing 16 models across 22 windows. Click legend items to hide/show series.

Recent Window Winners

Holdout Best Log Loss Runner-up Models
Apr 1 - Apr 6 Log Adjusted 0.537 Adjusted Efficiency (0.538) 16
Mar 25 - Mar 31 Log Adjusted 0.600 Adjusted Efficiency (0.601) 16
Mar 18 - Mar 24 Adjusted Efficiency 0.447 Log Adjusted (0.448) 16
Mar 11 - Mar 17 Margin Recency 0.599 Margin (0.604) 16
Mar 4 - Mar 10 Core Ensemble 0.587 Recency Ensemble (0.587) 16
Feb 25 - Mar 3 Points Off/Def Recency 0.578 Margin Recency (0.582) 16
Feb 18 - Feb 24 Margin 0.590 Points Off/Def (0.590) 16
Feb 11 - Feb 17 Recency Ensemble 0.614 Core Ensemble (0.615) 16

Model Performance Leaderboard

Models ranked by strict holdout AUC when available (fallback: full-season AUC). Hover over column headers for explanations.

# Model 7d Split AUC Acc Brier LogLoss n AUC 7d Acc 7d Brier 7d n 7d
1 Adjusted Context Blend Adjusted Context Blend Experimental context-heavy win model blending strong team components with rest and venue context. More → STRICT
13g
 - - - - 0 0.925 76.9% 0.180 13
2 Adjusted Efficiency Adjusted Efficiency Opponent-adjusted efficiency model with separate offensive and defensive components. More → STRICT
13g
 0.815 73.1% 0.179 0.529 5361 0.825 76.9% 0.179 13
3 Log Adjusted Log Adjusted Log-scale adjusted efficiency model that downweights blowout leverage. More → STRICT
13g
 0.814 73.2% 0.179 0.530 5361 0.825 76.9% 0.178 13
4 Points Off/Def Points Off/Def Raw points regression with separate offensive and defensive team parameters. More → STRICT
13g
 0.812 73.3% 0.182 0.541 5361 0.825 69.2% 0.196 13
5 Margin Margin Linear team-strength model fit on point differential instead of binary wins. More → STRICT
13g
 0.822 73.4% 0.178 0.531 5361 0.800 69.2% 0.195 13
6 Efficiency Efficiency Tempo-adjusted efficiency version of Pythagorean ratings. More → FULL
no 7d
 0.796 71.6% 0.191 0.571 5361 - - - 0
7 Margin Recency Margin Recency Margin regression with exponential recency weights on newer games. More → STRICT
13g
 - - - - 0 0.750 53.8% 0.225 13
8 Core Ensemble Core Ensemble Equal-logit blend of Elo, recency BT, recency margin, log-adjusted pyth, and points off/def. More → STRICT
13g
 - - - - 0 0.750 69.2% 0.215 13
9 Points Off/Def Recency Points Off/Def Recency Off/def points regression with exponential recency weights. More → STRICT
13g
 - - - - 0 0.675 61.5% 0.227 13
10 Recency Ensemble Recency Ensemble Equal-logit blend of Elo, recency BT, recency margin, log-adjusted pyth, and recency points off/def. More → STRICT
13g
 - - - - 0 0.650 61.5% 0.223 13
11 Home Team Baseline Home Team Baseline Always favor the home team with a fixed prior. More → STRICT
13g
 0.638 64.0% 0.232 0.657 5361 0.650 61.5% 0.237 13
12 Bradley-Terry Bradley-Terry Static logistic paired-comparison model with one team strength parameter. More → STRICT
13g
 0.823 73.8% 0.175 0.523 5361 0.625 61.5% 0.228 13
13 Dynamic Bradley-Terry Dynamic Bradley-Terry Time-evolving paired-comparison model with latent team strength drift. More → STRICT
13g
 - - - - 0 0.625 61.5% 0.248 13
14 Pythagorean Pythagorean Pythagorean win expectation from raw points scored and allowed. More → STRICT
13g
 0.752 69.8% 0.208 0.603 5361 0.600 61.5% 0.235 13
15 Avg Margin Baseline Avg Margin Baseline Predict from simple average scoring margin in the training window. More → STRICT
13g
 0.784 70.7% 0.195 0.573 5361 0.550 69.2% 0.239 13
16 Elo Elo Streaming paired-comparison rating with recency baked into sequential updates. More → STRICT
13g
 0.783 70.5% 0.191 0.564 5361 0.500 53.8% 0.263 13
17 Bradley-Terry Recency Bradley-Terry Recency Static Bradley-Terry with exponential recency weights on newer games. More → STRICT
13g
 - - - - 0 0.400 38.5% 0.266 13

Methodology

ELO / Bradley-Terry

  • ELO: Iterative updates, K=64, HCA=100
  • BT: Static logistic regression on all games
  • Both model win probability, not margin
  • ELO updates after each game; BT fits once

Pythagorean Models

  • Raw: Classic points scored/allowed formula
  • Efficiency: Pace-adjusted (pts per possession)
  • Adjusted: Opponent-adjusted efficiency
  • Log: Log-linear multiplicative scale

Margin Regression

  • Team-level ridge regression on point margin
  • Linear Bradley-Terry (margin target)
  • Alpha=0.05 (CV-tuned)
  • Learns home advantage from data (~6 pts)

Baselines

  • Home Team: Always predict home wins (60%)
  • Avg Margin: Higher average margin wins
  • Models should beat these to add value