🐻⬇️🏀

2025-2026 NCAAWD3 Model Performance Analysis

Comparing prediction accuracy across 4609 games using multiple rating models.

Model Performance Leaderboard

Models ranked by AUC on 2025-2026 games. Hover over column headers for explanations.

# Model AUC Acc Brier LogLoss n AUC 7d Acc 7d Brier 7d n 7d
1 Bradley-Terry 0.954 88.4% 0.129 0.429 1454 - - - 0
2 Margin Regression 0.954 88.0% 0.101 0.336 1454 - - - 0
3 Pythagorean Adjusted 0.943 86.2% 0.097 0.304 1454 - - - 0
4 Pythagorean Log 0.942 86.2% 0.098 0.307 1454 - - - 0
5 Pure ELO 0.909 81.7% 0.157 0.495 1454 - - - 0
6 Avg Margin (Baseline) 0.894 80.6% 0.134 0.416 1454 - - - 0
7 Pythagorean Raw 0.881 80.3% 0.145 0.446 1454 - - - 0
8 Home Team (Baseline) 0.555 55.5% 0.249 0.691 1454 - - - 0
9 Bradley-Terry Recency - - - - 0 - - - 0
10 Pythagorean Efficiency - - - - 0 - - - 0
11 Margin Recency - - - - 0 - - - 0
12 Points O/D - - - - 0 - - - 0
13 Points O/D Recency - - - - 0 - - - 0
14 Core Ensemble - - - - 0 - - - 0
15 Recency Ensemble - - - - 0 - - - 0
16 Adjusted Context Blend - - - - 0 - - - 0
17 Dynamic Bradley-Terry - - - - 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: Bill James formula (pts scored/allowed)
  • Efficiency: Pace-adjusted (pts per possession)
  • Adjusted: KenPom-style opponent adjustment
  • 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