🐻⬇️🏀

2025-2026 NCAAWD2 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

Comparing prediction accuracy across 1722 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 14 windows. Click legend items to hide/show series.

Recent Window Winners

Holdout Best Log Loss Runner-up Models
Feb 4 - Feb 8 Recency Ensemble 0.495 Margin Recency (0.496) 16
Jan 28 - Feb 3 Margin Recency 0.477 Margin (0.478) 16
Jan 21 - Jan 27 Margin Recency 0.476 Recency Ensemble (0.477) 16
Jan 14 - Jan 20 Recency Ensemble 0.487 Core Ensemble (0.487) 16
Jan 7 - Jan 13 Core Ensemble 0.462 Recency Ensemble (0.463) 16
Dec 31 - Jan 6 Dynamic Bradley-Terry 0.575 Margin Recency (0.581) 16
Dec 24 - Dec 30 Log Adjusted 0.055 Adjusted Efficiency (0.055) 16
Dec 17 - Dec 23 Recency Ensemble 0.514 Core Ensemble (0.515) 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 Margin Recency Margin Recency Margin regression with exponential recency weights on newer games. More → STRICT
290g
 - - - - 0 0.856 76.6% 0.159 290
2 Adjusted Context Blend Adjusted Context Blend Experimental context-heavy win model blending strong team components with rest and venue context. More → STRICT
290g
 - - - - 0 0.853 75.5% 0.159 290
3 Margin Margin Linear team-strength model fit on point differential instead of binary wins. More → STRICT
290g
 0.850 76.9% 0.161 0.489 2994 0.851 76.9% 0.161 290
4 Core Ensemble Core Ensemble Equal-logit blend of Elo, recency BT, recency margin, log-adjusted pyth, and points off/def. More → STRICT
290g
 - - - - 0 0.851 76.9% 0.158 290
5 Recency Ensemble Recency Ensemble Equal-logit blend of Elo, recency BT, recency margin, log-adjusted pyth, and recency points off/def. More → STRICT
290g
 - - - - 0 0.851 76.9% 0.158 290
6 Points Off/Def Recency Points Off/Def Recency Off/def points regression with exponential recency weights. More → STRICT
290g
 - - - - 0 0.848 76.2% 0.163 290
7 Points Off/Def Points Off/Def Raw points regression with separate offensive and defensive team parameters. More → STRICT
290g
 0.800 72.5% 0.184 0.548 2994 0.840 76.9% 0.165 290
8 Adjusted Efficiency Adjusted Efficiency Opponent-adjusted efficiency model with separate offensive and defensive components. More → STRICT
290g
 0.799 72.7% 0.195 0.620 2994 0.836 75.5% 0.174 290
9 Dynamic Bradley-Terry Dynamic Bradley-Terry Time-evolving paired-comparison model with latent team strength drift. More → STRICT
290g
 - - - - 0 0.835 75.9% 0.167 290
10 Log Adjusted Log Adjusted Log-scale adjusted efficiency model that downweights blowout leverage. More → STRICT
290g
 0.800 72.7% 0.194 0.616 2994 0.834 75.5% 0.175 290
11 Elo Elo Streaming paired-comparison rating with recency baked into sequential updates. More → STRICT
290g
 0.850 76.2% 0.161 0.492 2994 0.831 75.5% 0.169 290
12 Bradley-Terry Bradley-Terry Static logistic paired-comparison model with one team strength parameter. More → STRICT
290g
 0.861 76.7% 0.156 0.478 2994 0.828 77.6% 0.171 290
13 Bradley-Terry Recency Bradley-Terry Recency Static Bradley-Terry with exponential recency weights on newer games. More → STRICT
290g
 - - - - 0 0.828 74.5% 0.174 290
14 Avg Margin Baseline Avg Margin Baseline Predict from simple average scoring margin in the training window. More → STRICT
290g
 0.852 76.8% 0.160 0.485 2994 0.806 74.1% 0.180 290
15 Pythagorean Pythagorean Pythagorean win expectation from raw points scored and allowed. More → STRICT
290g
 0.790 71.9% 0.189 0.559 2994 0.730 70.7% 0.223 290
16 Home Team Baseline Home Team Baseline Always favor the home team with a fixed prior. More → STRICT
290g
 0.571 57.1% 0.246 0.685 2994 0.561 56.2% 0.248 290
- Efficiency Efficiency Tempo-adjusted efficiency version of Pythagorean ratings. More → FULL
no 7d
 - - - - 0 - - - 0

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