PUBLIC MODEL LEADERBOARD

Whose model is beating the baseline?

Opt-in user models scored against My CFB's public baseline on the exact same completed games.

Season
2026
Latest week included
Week 2
Latest included game
Sep 11, 2026

Log in to choose a public model

Snapshot published ; latest included evaluation recorded .

PUBLISHED STANDINGS

Model rankings

Every public model is ranked. The badge shows how much game history supports its score.

  1. #1
    my test v2NewcomerAnonymous Fan VG8XT8EUAD
    Score
    -0.36
    Eligible games
    97
    Straight-up
    84–13 (86.6%)
    Margin error
    25.0 pts Baseline 24.9
    Brier error
    0.114 Baseline 0.110
HOW SCORING WORKS

Same games, conservative credit

Lower prediction error wins

The score starts with the public baseline's mean Brier error minus the user's mean Brier error on the exact same games. A positive score means the user model performed better.

Small samples are restrained

Positive skill is multiplied by games / (games + 50). Negative skill is not pulled toward zero. This prevents a new model with one lucky result from jumping to the top.

Everyone is ranked

Every public model receives a numeric rank. Newcomers have 0–99 eligible games, Seasoned models have 100–499, and Veterans have 500 or more.

One model at a time

An account can change its public model twice per football week. After a switch, the selected model's own season record replaces the old model's record; the two records are never combined.

Only official pregame predictions with a matching public-baseline evaluation count. Predictions made before public enrollment count, but My CFB never creates or changes a prediction after kickoff. Tied games and missing comparisons are excluded. Model settings, individual predictions, selection history, and account details remain private.

Read the full model methodology