The MLB Underdog Edge: One Pocket, Measured Against the Sharp Line
A fixed MLB moneyline study measured a 4.2-point typical gap between Fairline's model and the Pinnacle close. One pocket stood out: model-favored underdogs beat the de-vigged close by 8.8 points over an April to June 2026 window (n=291). Here is that evidence, its limits, and the forward record testing it.
Updated Jul 2026 · Part of the models series
Fairline publishes one fixed sharp-line study and tracks one pocket from it. This page presents the measured result and its limits.
The study covers 855 MLB moneylines from April 12 through June 15, 2026. The model’s typical gap from the Pinnacle close was 4.2 points. The underdog cohort below produced a larger actual-minus-sharp result in that same window.
The pocket
The pocket covers MLB moneyline underdogs that clear two bars: a model win probability at least 3 points above the market’s own de-vigged probability, and a price of decimal 2.0 or longer.
Benchmarked against the de-vigged Pinnacle close over an April to June 2026 window, the pocket won 50.5% of its bets, against an implied 41.7% from the market. That gap is a +8.8 point beat-close on n=291, about three standard errors from zero. That window is a fixed retrospective study. The forward version is covered below.
Split by home and away, the beat-close stays positive on both sides. Home underdogs in the pocket beat the close by +9.6. Away underdogs beat it by +8.5.
The selection effect
Blindly betting every MLB moneyline underdog, no model filter at all, returned roughly breakeven results over the same window, +0.4% ROI against the de-vigged close.
The blind baseline controls for how all underdogs performed in the same study. The 3-point model filter separated the favored cohort from that baseline within this window. The forward record tests whether that separation persists on new games.
Every number in this comparison, the pocket’s +8.8, the blind baseline’s +0.4%, and the home and away splits, runs against the de-vigged closing line as its benchmark. Final scores supply the actual win rate and ROI. The +8.8 is actual results minus the sharp line’s implied win rate, not entry-to-close CLV. Closing line value covers why that benchmark reads faster than a plain win-loss record.
Where it stops
The +8.8 number compares actual wins with Pinnacle’s de-vigged close. The study also computed an 18.3% ROI at Pinnacle’s own price, which US bettors cannot take. It did not replay the 291 games at takeable prices, so neither figure establishes a return a US bettor could have earned.
The result shows up only on the current MLB evidence. Running the same selection method on NHL and EPL underdogs did not turn up the same pocket.
The sample is also young. The pocket covers a 291-bet window from roughly two months of games, and the record keeps testing it forward for signed-in members.
The live executable record uses takeable soft-book prices on flagged plays. It is a separate forward sample from the fixed study and reports its return separately. Pinnacle’s close remains the benchmark used to measure the result.
This is one fixed-window result. Any single game still turns on ordinary variance, no matter how the study reads. Variance in betting covers how much of a single result is noise.