Bullpens: the hidden variable in MLB predictions
Starters get the headlines, but games are decided in innings six through nine. How bullpen quality — and fatigue — moves win probabilities.
Everyone prices the starter. The market's blind spot sits behind him: modern starters average barely five-and-a-half innings, which hands a third of every game — usually the tensest third — to the bullpen.
Quality is priced. Fatigue isn't.
Season-long bullpen quality is public and mostly reflected in prices. What moves slower is usage: a bullpen that threw heavy innings three nights running is a different bullpen tonight, whatever the season numbers say. Managers know it — they hold gassed arms back — but the market often prices the name on the door, not the arms actually available.
How our model handles it
Bullpen quality is part of every rating our model produces, and per-recent-usage fatigue is the next refinement being wired into the daily board. It's a signal with the two properties we like most: it's quantifiable from public data, and it's boring — which is exactly why prices are slow to absorb it. The edges that persist are the ones too dull to tweet about.
Check any game on the MLB board: the starter line tells you where tonight's probability comes from — and when the model leans away from the market on a tired pen, the written read says so.
Frequently asked questions
How much of an MLB game do bullpens pitch?+
Modern starters average well under six innings, which leaves a third or more of every game — usually the highest-leverage third — in the bullpen's hands.
Every fixture, fully modelled — the correct-score grid, the derived markets, and the written read.
Start free