Andrew Knapp in game against the Miami Marlins

Excerpt of a report on explorable modeled estimates of catcher’s contributions to umpire calls and runs saved.

Explorable differences between umpire calls and modeled probabilities of strikes

Below, we can explore the percentage difference between the umpire calls and modeled probabilities of those calls, with corresponding video footage. Circles ○ in the top row represent called pitches (ball, strike) and lightness of the color represent how close¹ our modeled strike probability was to the actual call {0, 1} for each handedness (throw-stand) matchup {LL,LR,RL,RR}.

Of note, the graphics below only show pitches where the differences of called ball or strike from modeled probatility of strikes exceed ± 0.10.

Circles ○ in the bottom row represent the corresponding estimate of runs saved from that variation.

Hovering a pointer over a circle ○ links ○—○ pitches in top and bottom rows, and provides more details of the play in a tooltip. Clicking a pitch loads its game video:

Clicking one of the above observed pitches loads its video: