Next 24 hours
Best bets
Recommended bets first — the early spreads and totals that beat the market out of sample — then the model's most confident picks and biggest disagreements, for research.
Recommended bets
Early spreads & totals — the rule that returned +4.1% on 2,440 unseen bets last year1 unit each. Flat stakes — the backtest bet the same amount every time. A unit is ~1% of your bankroll.
Bet early. The edge comes from opening lines; it shrinks as the line moves toward these picks.
Shop the price. The best price we see is shown with its book. Skip a bet if the line has moved past it.
Backtested, not guaranteed: the +4.1% comes from last season's opening lines, and it is being verified live on the Performance page. Expect long losing streaks even with a real edge. 21+ only; bet only what you can afford to lose.
The sections below are research, not recommendations. “Best” means the model's strongest view, not a guaranteed edge. The confidence picks are well calibrated — when the model says 80%, it has historically won about 80% of the time — but favourites pay little. Market disagreement has not been backtested yet. Use these as research starting points.
Strongest model picks
Most confident win probabilitiesValue vs market
Model probability above the vig-free market probability, ranked by model expected value per unit — research candidatesΔ = model − vig-free market probability (proportional vig removal). Model EV assumes the model probability is exactly right; the track record below tests that assumption. Spreads and totals use the Monte-Carlo score distribution conditioned on no push.