Sports analytics you're allowed to check our work on
Most handicapping services sell you a pick and a promise. Baseball AI Tools is built the other way around: a documented scoring engine, real data sources, and a public track record — win or lose.
Fundamentals, weighted transparently
Recent form over the last 10 games (recency-weighted, so last night matters more than three weeks ago), starting pitcher trend/rest/fatigue, bullpen workload, lefty/righty platoon edges, park factors, and game-time weather all feed a documented scoring formula — not a prompt to a language model.
Technical indicators borrowed from trading
The same primitives quants use on stock price series — EWMA, rate-of-change momentum, an RSI-style hot/cold oscillator, and Bollinger-band-style volatility — applied to each team's run-scoring series to catch streaks and regressions the raw stat line misses.
Graded, not just published
Every pick is checked against the actual final score once the game ends. Side and total results are recorded, win or lose, and feed directly into the next principle.
A backtester that proposes, never auto-changes
Periodically, the backtester measures which highlight categories actually correlated with wins and proposes nudged model weights. Nothing goes live automatically — a human reviews and activates each change.
Where the data actually comes from
Schedules, boxscores, team and pitcher stats, and roster/handedness all come from the free MLB Stats API. Game-time weather comes from Open-Meteo. Sportsbook lines come from The Odds API, snapshotted a few times a day to track how the line moves. The one honest gap: there's no free, machine-readable feed of daily umpire assignments or zone tendencies, so the engine uses a neutral factor until that data is available — we'd rather tell you that than fake it.
What "line movement" is — and isn't
When a pick's line has moved since our first look at it today, we say so. That's a real, honest signal: raw market movement is something bettors do watch as a rough tell for sharp/steam action. What we don't have — and won't pretend to have — is bet-percentage or handle data, so we can't tell you whether a move was driven by sharp money or a public pile-on. "Line Movement" measures the move itself, not who caused it.