EDGEModel Performance
Every graded pick the model has made, checked against real final results — not just the ones anyone happened to bet on. Two-way markets are counted once, on the side the model actually favored.
By Grade
Only picks with a real book price get a grade — it's computed from edge vs. that price. That's most Moneyline, Run Line and Total picks, plus a large share of Team Total and First 5 picks; genuinely unpriced picks appear in the by-market table below without a grade. Same recommended-side counting as the overall record above, so these rows add up to it.
The letter grade reflects how decisive the model's own probability looked when the pick was made — NOT whether picks like it have actually won. The badge next to each tier is the honest answer to that: Proven means this tier's real win rate clears the 52.4% breakeven bar even under a conservative statistical read (95% Wilson lower bound) with at least 30 decided picks — Not enough data means it hasn't reached that sample size yet, and Underperforming means it has, and doesn't clear the bar.
By Market
Win rate of the side the model actually favored (higher of the two probabilities per game), not a raw count across both sides — a market with two-way projections (e.g. both teams' moneyline) would otherwise always land near 50% regardless of model quality, since exactly one side wins every settled game. Includes model-only markets (no live price), shown for transparency even without a grade attached.
Previous Day
Yesterday's graded picks broken down the same way as the all-time record above: by letter grade and by market, so you can see how the most recent slate did on its own without it being buried in the full history.