Help improve Stats Over Score! Report bugs, share feedback, or ask us about anything you see: [email protected]
because the score is just another stat anyway
Advanced MLB predictive model for strikeouts, walks, hits, and runs
⚾ Stats Over Score: MLB — A Work In Progress. Please Excuse and Report any Bugs!
What does SOS Premium Entail?

Track Record

Every projection graded against the box score the next morning — misses included.

How the model is doing
Hit calibration — when we say it, does it happen?
Full report — every graded dimension

Graded slate
How grading works

Every stat is graded the way its distribution deserves

Pitcher lines (K, BB, hits, earned runs) and game totals are point projections with meaningful magnitudes, so they're graded on average miss (MAE), within-a-threshold rates, and bias. Batter counting stats have low per-game means — a batter projected at 0.83 hits going 0-for-4 is the statistically expected outcome about 40% of the time, not a miss — so hits, batter strikeouts, and walks are graded on calibration: the probability our projection implies versus how often the outcome actually happened, across thousands of batter-games. HRR (hits + runs + RBI) carries a high enough mean to grade on average miss directly.

Nothing is cherry-picked

Every projected starter who took the mound is graded. Scratched starters are excluded (we didn't project the guy who actually pitched) but counted and shown. Injury-shortened outings stay in the record, flagged. The biggest miss gets the same billing as the best call.

Coverage notes