Every baseball fan knows the feeling: it’s the 8th inning of a tight ballgame at Fenway, the manager strolls out to the mound, and the TV broadcast puts up a graphic showing the reliever’s pristine 2.40 ERA.
Thirty seconds later, he gives up a 420-foot missile off the back wall of the bleachers.
The problem with standard box scores and traditional ERA is that they don’t capture cumulative strain. Throwing 15 high-stress pitches on three consecutive nights puts vastly more wear on an arm than throwing 45 pitches once and resting for three days.
To see who was actually fresh before first pitch every morning, I built Dirty Water, an automated daily analytics dashboard tracking Statcast pitch physics, bullpen availability, and Fenway matchup geometry.
The 06:00 ET Wake-Up Call
The pipeline runs on autopilot every morning. A Python cron job queries MLB’s Statcast API for the previous night’s pitch-level tracking data:
-
Release mechanics: Initial velocity (v₀), release point coordinates, and induced movement (IVB & horizontal break).
-
Batted-ball quality: Exit velocities and launch angles converted into expected metrics (xBA, xwOBA).
From there, it computes an Availability Score (0 to 100) for every arm in the Boston bullpen. Instead of treating pitches linearly, the model applies exponential decay weights over a rolling 4-day window and adds metabolic penalties for high-leverage innings with runners in scoring position.
When a reliever drops below an availability score of 40, opponents’ hard-hit rates against their breaking balls jump by over 6%, and walk rates spike. It turns out arms really do get tired, even when their season ERA looks great on TV.
Fenway Is a Weird Place
The other half of the tool is matchup simulation. Fenway Park’s outfield dimensions are famously bizarre, from the 310-foot Green Monster looming in shallow left to the cavernous 420-foot “triangle” in center.
By mapping projected starting pitcher repertoires (sinker/sweeper/cutter splits) against opposing batter spray charts, Dirty Water simulates which hitters are primed to exploit Fenway’s quirks on any given afternoon.
The whole dashboard is built in Python using Plotly and Dash, containerized in Docker, and live at dirtywater.corygarms.com.