2026 MLB Postseason Bracket and Engine Report

How to read this table: the engine pick is the side our 1000-sim Monte Carlo favored; the calibrated column applies the production home-axis shift map fitted through 2025. The market column is the sportsbook moneyline captured at lock time, de-vigged to a probability. When the engine and the market disagreed by ten points or more, our shade rule called NO-BET and we claimed no pick: after an engine-read concluded our simulated signal is a strict subset of the information already priced into the market, a double-digit disagreement is treated as the engine being wrong, not as edge. That rule went 4-for-4: in all 4 flagged games the engine side lost and the market side won, including ALCS Game 6, where the engine favored Cleveland at 61.5% raw while the market favored Tampa Bay, and the Rays closed it out 7-3 to win the pennant.

On the games where the disagreement was small enough to claim a pick (3 games), the engine went 1 for 3 and the market went 1 for 3. That is not a resume; it is the sample. The point of publishing it is the discipline: the same table that shows the engine nailing ALCS Game 3 (Cleveland, 58%) also shows it whiffing on both Milwaukee wins in Los Angeles. We do not shade that. See the methodology for the calibration details and the player projections pages for the same honesty applied to season-long forecasts.

Results above are from the official MLB Stats API postseason schedule. The engine-vs-market ledger covers only the games we actually simulated before first pitch, starting October 13; earlier rounds are shown as results only, which is itself the honest framing: we publish probabilities when we have receipts for them, not retroactively.