How Do Starting Pitchers and Bullpen Fatigue Change a Projected Run Total?
The MLB model splits a game into starter innings and bullpen innings, then adds 0.3 to 0.5 runs when a bullpen is stretched. Here is the arithmetic.
Updated Sep 2026 · Part of the models series
The model prices a baseball game in two pieces. It rates the starting pitcher across the innings he is expected to throw, rates the bullpen across the innings behind him, and adds the two. A stretched bullpen then adds a flat 0.3 to 0.5 runs to whatever the other team is projected to score.
How does the model split a game between the starter and the bullpen?
It expects 5.5 innings from a starter and 3.5 from the bullpen. Each segment carries its own rating, and the projected runs for a team are the two segments put together.
The starter’s rating blends three fielding-independent measures, FIP, xFIP and SIERA, across three time windows: his last three to five starts, his current season, and his prior year. The bullpen’s rating comes from the unit’s expected fielding-independent line against the league average. Both come out as a factor rather than a run total, which is what lets the model weight them by innings and add them. When a team announces an opener instead of a starter, the starter’s share of the innings shrinks and the balance moves to the bullpen.
What counts as a tired bullpen?
Two things flip the flag. Either the team’s relievers threw at least 11 innings across the previous two games, or at least three individual relievers each carry a fatigue flag of their own.
Eleven innings over two games is 5.5 innings a night from the pen, which is a genuinely stretched unit. The threshold was 8.0 innings until 2026-05-06. At 8 it fired on ordinary usage, because starters average around 5 innings and a 4-inning-a-night pen is normal. A review across 560 team-games found the old threshold firing for 73% of teams and adding about 0.59 runs a game across the board.
A single reliever picks up a flag when he threw 2 or more innings the previous day, made 4 or more appearances in the last four days, or threw 30 or more pitches in his last outing. In an eight-man bullpen, three of those at once means the pen is depleted. An off day clears everything and resets the team to fully fresh.
What is the fatigue adjustment worth in runs?
Between 0.3 and 0.5 runs, sliding with how depleted the pen is. The model scores the unit on a freshness value between 0 and 1, then reads the adjustment off that.
Order matters here. The multiplicative stack of pitcher, lineup and environment is shrunk toward the league average first. Home field, a velocity decline penalty and this bullpen fatigue figure are added after that shrinkage, so the full 0.42 lands on the number undamped.
Why is first five innings a cleaner read on the starter?
A first-five-innings bet settles before the bullpen throws a pitch, so none of the fatigue arithmetic touches it. The starter rating is the model’s sharpest input, and this is the market that isolates it.
Modern starters average 5.0 to 5.5 innings, so the bullpen decides a large share of any full-game result. When one team has a strong starter in front of a stretched pen, the full-game price and the first-five price are answering different questions.
What this does not tell you
The fatigue flag covers the whole unit, not a plan for who pitches the eighth. It cannot know that the closer is unavailable while the rest of the pen is rested. The 5.5 and 3.5 innings are expectations, not a script, so an early hook moves the real split away from what the model priced. Every threshold on this page is refitted from results and has moved before, so read the current values off the MLB model page, which renders them from the running system. None of this reads a sportsbook line, and none of it makes a game a bet.