Grading the decisions,
not the players.
Every in-game decision has a correct answer in expected-value terms. We find it, grade the actual call, and surface who's adding or costing their team runs across every decision type we can measure.
How it works
Every pitch, baserunning opportunity, steal attempt, or tactical call is a discrete, time-stamped decision with a clear binary or categorical outcome.
RE24 run-expectancy tables give us the exact threshold at which a risky call (or a swing) is worth it in expected-runs terms, before any outcome is known.
We estimate true probability using empirical binning on observable pre-play factors. Good decisions beat break-even; bad ones don't.
Pitch decisions graded
4.50M
ESV module alone, 2020–2026
Seasons covered
7
2020–2026 · 2026 season complete
Live modules
4
ESV · Send/Hold · Pinch Hit · Steal
Avg YoY rho
+0.60
ESV RAA/100, hitters with 750+ pitches in both seasons, 2021–2026
Expected Swing Value
For every pitch a batter faces, there is a correct decision: swing or take. We compute the Expected Swing Value (ESV), the difference in expected run value between swinging and taking, using the pitch type, location, velocity, count state, base-out situation, and the batter's own contact history against that pitch profile. Grade the actual decision against the higher-EV option and aggregate to a season leaderboard showing who is adding (or costing) their team runs through pitch-selection quality alone.
More decision modules
Each module grades a specific category of in-game call using the same expected-value framework.
Send/Hold Grader
Grading third-base coach decisions on extra-base opportunities
2020–2026
Pinch Hit Grader
Did the pinch hit substitution improve expected offense given the pitcher matchup?
2020–2026
Starter Pull Timing Grader
Scoped for a future release
Bunt Decision Grader
Scoped for a future release
Also available:
Scope and limitations
This tool grades the decision layer only: whether a call was correct given what was knowable at the moment it was made. It does not measure player ability, execution quality, or factors outside the decision-maker's direct control. All probability estimates use empirical binning on season-average statistics; play-level tracking is not available for every input. 2020 entries reflect a 60-game season and carry higher uncertainty. Full methodology →