No strong model edges today.
V10 passes when the model and the market agree — most games.
V10 is a logistic regression model trained on 7,330 MLB games from 2023–2026, validated at 55.9% honest out-of-sample accuracy. It finds games where the model's win probability differs from the Vegas-implied probability. That difference is what the model flags — not a claimed edge over the market itself.
Plays are the model's genuine picks: high win probability, positive expected value, and no contradicting market signal. Leans are moderate-confidence, lower-conviction calls. Everything else is a Pass — most games, where the model has no clear read.
Every prediction is published before first pitch. Every result is tracked. Nothing is sold. This is analytics research, not gambling advice.