Conservative by Design
A six-year calibration audit of DataGolf's prediction model — 292 events, 36,656 player-event records. The model is well-calibrated overall, but it systematically underestimates elite players by 5+ points.
Read the study →
The Market Already Knows
Measuring DataGolf's edge against Pinnacle's closing line across 284 events. The win market is fully efficient — but the less-liquid top-10 market hasn't caught up to the model.
Read the study →
Same Number, Higher Price
Eight sportsbooks, one signal. Recreational books price as accurately as Pinnacle — they just charge roughly double for it. And the "popularity premium" everyone assumes exists runs backwards.
Read the study →The Model Moves the Market
Seven seasons of opening and closing prices. The model’s picks shorten before the market closes — every market, every threshold, 48 out of 48 measurements. Whether that edge clears the hold is a different question, and the honest answer is that seven seasons cannot yet separate them.
Read the study →