MLB predictions today: how our pitcher-aware model works
Modal's MLB model fuses team-strength Elo with the starting pitcher's FIP, then blends against the market. Here's the full method behind our daily MLB win probabilities.
Baseball punishes lazy models. A daily slate, thirty teams, and a market sharpened by decades of professional money — if your method is vague, the market eats it. So here's ours, in full.
Team strength: an Elo that never sleeps
Every team carries an Elo rating updated after every game — win against a strong side and you gain more than beating a weak one. Elo's virtue is honesty: it doesn't care about narratives, only results, weighted by opponent quality. Ours is fit on multiple seasons of games and re-tuned on a walk-forward basis, so the hyperparameters are chosen by predictive power, not by taste.
The starter: baseball's single biggest variable
A baseball game is not team versus team — it's lineup versus starting pitcher, twice. Our model rates every probable starter, weighted first by FIP (Fielding Independent Pitching), which counts only what the pitcher controls: strikeouts, walks, hit batters, home runs. ERA gets polluted by the defence behind a pitcher; FIP doesn't. The bullpen behind the starter is priced in as well.
The blend: respecting the market
Here's the part most model-sellers won't tell you: months of testing showed the sharp market consensus was more accurate than our raw model — and that the blend of the two beat both. So our published betting probability is deliberately a fusion: mostly market, corrected by model. Pretending your model alone beats Vegas is marketing; measuring where it adds signal is method.
Sorted by disagreement
The MLB board shows every game on today's slate — model, market and blended probability, both starters with their FIP — sorted by how far the model sits from the market. The top of the board is where the model thinks the price is wrong. That's where the reading is, and where the engine behind Momus's real-money record places its own bets.
Free during the beta. Analysis, not tips — you make your own decisions.
Frequently asked questions
How accurate are Modal's MLB predictions?+
In shadow testing across 79 graded bets against sharp closing prices, the engine hit 59.5% with a +6.3% return — and the model's straight-up game picks currently beat the market consensus on accuracy across 370+ graded games. All numbers update publicly as games settle.
Why does the model care so much about starting pitchers?+
No single player moves a baseball game's probability like the starter. Two otherwise identical teams can swing several points of win probability purely on who takes the mound — which is why our team rating is fused with a starter rating on every game.
What is FIP in baseball?+
Fielding Independent Pitching measures only what a pitcher controls — strikeouts, walks, hit batters and home runs — stripping out fielding luck. It predicts future performance better than ERA, which is why our model weights it first.
Every fixture, fully modelled — the correct-score grid, the derived markets, and the written read.
Start free