Pitcher fatigue is the measurable decline in velocity, command, and outcomes once a starting pitcher has crossed his individual workload threshold. For most modern starters, that's 85-95 pitches or the third time through the order, whichever lands first. The collapse, when it happens, lives in innings 5-7. Books anchor the total to the starter's season ERA, which dilutes the late-game collapse risk. That gap is where the edge sits.
I'll cover the two markers the Bookie Bullies model actually uses to flag a fatigued starter, the math on why they create an over/under edge, the bullpen variable that decides which way the edge goes, and the Quality Start trap that misleads casual bettors into the wrong side of a fatigue collapse.
The cleanest fatigue signal is fastball velocity. Most starters lose 1-2 mph between pitch 70 and pitch 100. The drop is visible on Statcast pitch-by-pitch in any live broadcast, but the actionable signal isn't the in-game decay. It's the year-over-year first-inning velocity delta.
A pitcher who threw 95 mph average in inning 1 last year and is now sitting 94.2 has lost 0.8 mph at his freshest moment of every start. His pitch-100 velocity floor, by extension, is lower than his pitch-100 floor a year ago. That cumulative loss compounds with the in-game decay curve. The model treats year-over-year velocity drop above 0.8 mph as a multiplier on standard fatigue risk, not an additive factor.
This is why a 33-year-old veteran with a "good ERA" can still be the under play. The season ERA reflects his average outing. The model is asking what his inning-7 outing looks like. Different question.
Batters hit about 30 points of OPS better the third time they face a starter compared to the first. The reasons are debated. Familiarity probably matters most. Cumulative fatigue stacks on top of it. Manager reluctance to pull a starter who hasn't pitched poorly yet is the structural part.
The effect is real, consistent across 20+ years of MLB play-by-play data, and well-priced by analytics-driven managers. Most front offices now pull starters before the third TTTO. The exception is when the bullpen can't cover. If a team's high-leverage relievers are tapped out from the previous two games, the manager has to ride the starter into his third TTTO. That's the moment.
I wrote about how to track 72-hour bullpen workload last week. The pitcher-fatigue edge and the bullpen-depletion edge are the same edge, viewed from opposite sides of the same game.
Sportsbooks price MLB totals primarily off two inputs: the starter's season-long ERA and the lineup's wOBA against same-handed pitching. Modern books layer bullpen quality on top, but the bullpen factor anchors to a season-long bullpen ERA, not the day-of availability.
That leaves a gap. The book has the starter's first two times through the order correctly priced. It has the bullpen's average performance correctly priced. What it misses is the fatigue-window transition: innings 5-7 when the starter is too tired and the bullpen isn't ready to take over. Those innings are between the two pricing regimes.
A Quality Start is 6 innings of 3 earned runs or fewer. Casual bettors read it as "good pitcher outing" and lean under. They miss the structural problem: a starter who gives up 3 runs across 6 innings often gives up most of them in the fatigue window. Their final line looks acceptable. The middle innings looked terrible.
A typical example: starter throws 5 shutout innings, gives up 3 runs in the 6th, gets pulled. Final line is 5.2 IP, 3 ER. That's a Quality Start. The over hit in the 6th and the game ended 7-4. The "good outing" line on next week's preview will be a trap if the same fatigue markers are present.
If you're betting unders on a guy with a 3.50 ERA and 80% Quality Start rate, you're betting the average. You're not betting the marginal start. The marginal start with both fatigue markers active is the over.
Pull the starter's last 3 Game Log entries. For each: average fastball velocity, pitch count when pulled, and whether they faced the lineup a third time. If the velocity is trending down across the 3 starts and they've been pulled before TTTO in 2 of 3, the bullpen workload (see the previous post) is your conditional variable. Fresh bullpen behind them, the under can still win. Tired bullpen behind them, lean over.
Cross-reference against the day's posted total. If the book is pricing on the season ERA only (the line is roughly 0.3 runs below or above the matchup's true expected total), the edge is in the gap. If the line has already adjusted (visible by the move from morning to first-pitch in the line history), the edge is gone.
The cleanest pattern: a starter with a 3.40 season ERA, year-over-year velocity down 0.9 mph, last 3 starts pulled at exactly 6 innings (efficient enough to clear TTTO once but not twice). His opponent's lineup has a top-5 wOBA against his handedness. The book sets the total at 8.5 anchored on the starter's season ERA. The model projects 9.4 once the fatigue adjustment fires in innings 6-7. The over is 60-40, the book has it at 50-50.
The catch: if both teams' starters fit this profile, both fatigue windows hit, and the total moves to 9.5 by first pitch. The morning is the only window with the edge. By 4 PM PT, sharp money has confirmed the pattern and the line catches up.