Public model evidence
Model Leaderboard
Compare verified public model versions. Choose a market and the evidence that should determine rank.
Market
What do you want to compare?
MLBTotal0 rankedPolicy mlb-public-v130 graded / 30 metric minimum
Log loss leaderboard
Probability accuracy with a larger penalty for confident forecasts that are wrong. Lower is better; 0 is perfect.
Quick read: No model version currently has a ranked Log loss result.
empty
No Total public evidence yet
No public model version currently has recorded total evidence available for this leaderboard. Choose another market or browse the public models. No rankings are being withheld.
Method, sample rules, and exclusions
Policy mlb-public-v1. A version needs at least 30 graded forecasts and 30 observations eligible for Log loss. Lower is better. Each market and immutable version is evaluated separately. Rankings never include backtests, private-era forecasts, research results, or user decisions. Voids and invalidations remain visible in each version’s record but are excluded from metric calculations under the shared public-performance contract. Pausing or ending tracking stops future collection; it does not remove or rerank evidence already accumulated.
Rank is the beginning of the question
Compare sample, period, uncertainty, and evidence—not only position. Then test a model whose assumptions you can inspect.