Bullpen usage is the count of pitches and innings each relief pitcher has thrown in the previous 24, 48, and 72 hours. When a manager's high-leverage relievers are tapped out, the back end of his bullpen handles innings 7-9 instead. That single substitution swings late-game run rates more than starter quality, weather, or park factors for that one game. Most bettors don't track it. Sharp models do.
A reliever's effectiveness drops after consecutive-day appearances. Velocity dips 0.5-1.5 mph the day after a high-pitch outing. Command degrades faster than velocity. The recovery curve for most arms is 48 hours; for closers, who throw with maximum effort on every pitch, it's closer to 72.
The Bookie Bullies model checks "can this team's bullpen cover innings 7-9 with their A-tier arms?" on every game, every morning. It's not just asking who pitched yesterday. It's checking the rolling 72-hour pitch count for every name in the bullpen, weighted by leverage tier. A setup man burning through 25 pitches in a 1-run game counts more than a mop-up reliever throwing the same in a blowout.
The cutoff: if a team's top three high-leverage relievers have 4+ appearances combined in the last 72 hours, the bullpen flags as depleted for the next 24 hours. That triggers an adjustment on the game total.
The math is simple. An A-tier bullpen averages around 3.20 ERA. The B-tier averages 5.40. In a typical 9-inning game, those three relievers cover innings 7-9, which is 3 innings of work.
A 2.20 ERA gap over 3 innings is 0.73 expected runs of difference. If both teams' bullpens are depleted, the swing is roughly 1.5 runs across the full game. At a 9.5 total, 1.5 runs is the difference between the under being a 60% proposition and a 38% proposition. That's a 22-point implied-probability shift, which is what creates the edge before the line catches up.
Pull the day's slate. For each game, identify both team's top three relievers (the 7th-inning specialist, the 8th-inning setup man, the closer). Check their pitch counts and appearances over the previous 72 hours via any public box-score archive. If 4+ of the 6 names have pitched in the previous two days, the back-end bullpen is going to handle late innings.
Cross-reference against the day's total. The gap between what the line implies and what the bullpen quality projects is the bet. The Bookie Bullies free MLB picks card runs this calculation every morning and flags games where the depleted-bullpen signal contributes at least 0.6 runs of model edge. Those are the picks most likely to move totals.
Closers usually throw one inning when their team is ahead in the 9th. That's it. They're not the bottleneck. The bottleneck is the 7th and 8th inning guys, who pitch in more games and burn out faster. A closer who breezed through a 1-2-3 inning yesterday is fully rested today. A setup man who threw 25 pitches in a tight 8th is not, even though most casual bettors call him "fresh" the next morning because his name wasn't in the save column.
A 10-inning game burns 5 relievers, not 3. If yesterday's box score shows a 10-or-11-inning final, the entire bullpen pipeline got pulled forward. Today the rookies and journeymen pitch the late innings. Over moves regardless of starter matchup. (And if both teams played 10+ yesterday, you're looking at a near-guaranteed mispriced total this morning.)
Bullpens recover slowly. The 72-hour window matters more than the calendar day. A team coming off an off-day where both starters went 7+ innings has a fully rested bullpen. A team off a day where the starter went 4 innings does not, even after the off-day. The arms that pitched the most in the previous high-pitch game are still down a tick on velocity and command 48 hours later. I see "off-day = fresh bullpen" assumption all the time. It's worth maybe 30% of the recovery the public thinks it is.
Most days, two games on the MLB slate trigger the depleted-bullpen flag. The cleanest pattern: a team that just played 11 innings yesterday hosts a team coming off a normal 9-inning game where their starter went 7. The home bullpen is gassed. The visiting bullpen is fine. The model projects roughly 1.2 extra runs for the visiting offense in innings 7-9, which translates to about half a run of edge on the over.
The morning line, set on starter matchups alone, doesn't reflect this. By 4 PM PT, sharp money has confirmed the bullpen availability and the line moves. By 5 PM PT, the edge is gone. The whole game is publishing the pick in the morning window, before the line catches up.
That's why we publish twice daily, with the morning run anchored at 8 AM PT. The afternoon run reprices everything once lineups confirm. The public track record shows the hit rate on those morning-window total picks specifically.