04 · Research
The desk doesn't sell picks. Every game, the model logs a home-win probability before first pitch; we score it against the outcome and against the no-vig sharp line. This is the honest scoreboard — it shows the model tracks the market closely and has no standalone edge. It's here for transparency, not as a bet signal.
Brier scoremeasures probability accuracy — lower is better, and a model with a real edge would beat the market's Brier. Ours sits right on top of it. That's the point: the model is a market-grade read, not an oracle. Agreement is how often the model and the no-vig line favor the same side.
When the model says 60%, does the home team win ~60% of the time? Each bar is a probability bucket (season to date): the marker is what the model predicted on average, the bar is what actually happened. Close alignment means the probabilities mean what they say.