True odds on every side of every game, with the bookmaker’s cut stripped out, so you can see what you are really being offered and what the house is charging you for it. One bet in each game is picked and graded. Tap any bet for the full breakdown.

Rows are the four bets every game settles. Columns are how our model's number compared with the market's on that bet. Colour is return on a flat 1-unit bet, not accuracy. Cells under the sample floor are greyed: an ROI on twenty bets is decoration. Live and backtest are never blended — every backtest day was predicted by a model fit on data that already contained its outcome, so its edge is inflated by construction.

A unit is one flat bet. Winning at −110 returns 0.91u, losing costs 1.00u, so break-even is 52.38%. Accuracy is not money. −110 is a fiction for this model. It picks favourites at a median closing −139, where break-even is 58.1%, so the −110 column flatters it by a wide margin. The right-hand column prices every pick at what it actually closed at.

Disjoint bands, so each row stands alone. The cumulative tiers below nest every high-confidence pick inside every lower one, which flatters the weaker tiers.