The engine plays baseball the way baseball happens. Duel by duel, then it tells you why.
Baseball is a chain of pitcher-versus-hitter duels, so the simulation is the most literal in any KaNeXT sport: it runs the plate appearances that swing the game, over batting-order turnover, and reads out the drivers, not just a score. It prices pitcher fatigue and the times-through-the-order penalty, deploys the bullpen by leverage, carries the variance of a short series honestly, and rides the park and run environment through every level. Amateur and pro games simulate in their own engines, with no cross-engine mixing, and every win probability and projection arrives with its variance and confidence attached.
A win probability that names its duels.
The engine loads the pitcher-versus-hitter interactions for the matchups that swing the game, the starter and the leverage relievers against the heart of the order, by handedness and count, and runs the plate-appearance-and-inning model over batting-order turnover. The output is a win probability with the key matchup drivers, not a bare number.
The 58% home win probability is the sum of the duels that produced it, and the biggest single mover is the ace against the left-handed three-hitter in the highest-leverage spot. That is the most literal simulation in any KaNeXT sport, because the sport itself is a chain of pitcher-versus-hitter duels. A win probability that names its duels, not a score handed down.
Illustrative engine read on the real single-game plate-appearance model (pitcher-versus-hitter interactions by handedness and count, over batting-order turnover, decomposed to the leverage duels). Composite clubs, demonstration figures.
Fatigue is a number. The bullpen turn is win probability.
Pitcher fatigue and the times-through-the-order penalty are modeled explicitly: the starter's effectiveness decays as a lineup sees him a third time, and the engine prices the exact moment the game turns from the starter to the bullpen. Then it deploys the bullpen by leverage, because the highest-leverage innings should draw the best arms.
The starter is a .290 pitcher the first time through and a .348 pitcher the third, and the engine prices the 6th-inning hinge and the leverage-weighted bullpen behind it, worth about 5% of win probability between the actual and the optimal move. Fatigue and the third time through are not flavor; they are the model. The third time through is a number, and the bullpen turn is worth win probability the box score never shows.
Illustrative engine read on the real times-through-the-order penalty, pitcher fatigue, and leverage-weighted bullpen deployment (the decay curve, the turn point, the actual-versus-optimal win-probability delta). Composite clubs, demonstration figures.
Run it many times, and carry the variance honestly.
Extend the plate-appearance model across a series or a season: run the matchups repeatedly, carry the variance, and output win probability, playoff odds, or a projected record with an honest band, because a short series is high-variance and the confidence must say so. The park and run environment ride through every level.
The five-game series is a 55% home edge inside a wide 41-to-68 band, the season a projected 89-73 inside a 12-win band, and the same two teams post different odds in a hitter-friendly park than a pitcher-friendly one. The variance is carried, not hidden, because a short series is exactly where a false point estimate would lie to you. The shorter the sample, the wider the band, and the simulation refuses to pretend otherwise.
Illustrative engine read on the real series-and-season projection (repeated matchups, carried variance, win probability and playoff odds with honest bands, the park and run environment riding through). Composite clubs, demonstration figures.
The engine simulates baseball the way baseball happens, duel by duel, and then tells you which duels decided it. It prices fatigue and the third time through, deploys the bullpen by leverage, carries the variance of a short series honestly, and stays weightless: it plays out what is likely and traces why, and it does not tell you which game was supposed to matter. A win probability that names its duels is a bet you can interrogate; a bare number is a verdict you have to trust.
Run the duels. Read the drivers.
Simulation plays the game out plate appearance by plate appearance, decomposes the win probability to the matchups that produced it, prices the third time through and the leverage bullpen, and carries the variance of a series or a season with an honest band.