University Prediction Profiles - AI March Madness 2026

Per-university prediction profiles showing how each AI model rates every team in the 2026 NCAA Tournament. Tracks the number of times each team was picked to win, average confidence score per team per model, and source citation frequency.

University profiles reveal model-specific team preferences. A model that consistently picks a team with high confidence across multiple games may have strong evidence for that team's quality, or may be over-indexing on brand recognition and historical program prestige.

Mentions per game tracks how often each team appears in cited sources across all models. Teams with high media mentions tend to receive higher confidence scores from GPT-4o (which relies on major sports media), while analytically strong but under-covered teams may be better identified by Perplexity.

Team Pick Rate

The percentage of games involving a team where the model predicted that team to win.

Average Confidence

Mean confidence score across all predictions involving a specific team, by model.

Mentions Per Game

How frequently a team appears in cited source URLs across all models' prediction responses.