Bullpen Usage Metrics MLB Bettors Should Track

Bullpen usage is the record of who a team's relievers are, how good they are underneath their ERA, and how much they have thrown in the last few days. This page is for bettors who play MLB moneylines, run lines and totals, or player props that depend on late innings, and who want a repeatable way to read a bullpen before the market does.

The common mistake is to treat bullpen quality as the edge. It isn't, or not by itself. VSiN's Steve Makinen tracked a system that simply backed the team with the better-rated bullpen in every game of the 2026 regular season. It won 53.3% of the time, finishing 1113-975, and still lost 49.61 units for a -2.4% ROI, according to VSiN's final 2026 bullpen systems update. The market already knows which pens are good. What it prices less carefully is which pens are tired, which are trending, and which numbers are luck. That is where the work goes.

Relievers now pitch a large share of every game, so the bullpen matters more to a full-game bet than it used to. Baseball Bet of the Day's bullpen guide (updated October 2026) puts the reliever share at 40% or more of all innings, with starters averaging under six innings a start. ParlayTools puts it at nearly 45% for 2026. Either way, close to half of every full-game moneyline, run line and total is decided by pitchers who aren't named on the betting board.

The argument for paying attention to usage comes from how lines are built. Makinen's postseason trends piece for VSiN argues that oddsmakers still price games mainly on lineups and starting pitchers, which leaves bullpen fatigue as the input most likely to be underpriced. That fits the regular-season evidence above. A bullpen's quality is visible in every season table, and it's baked into the number. A bullpen's freshness changes daily and has to be counted by hand.

So there are three ways to read a bullpen, and they aren't equally useful. Results are what happened. Skill metrics are what is likely to keep happening. Workload is what the pen can do tonight.

The best evidence that skill plus workload actually predicts comes from Travis Sawchik's stretch-run model on MLB.com. It combined K-BB%, SIERA, FIP, Stuff+ and Location+ with freshness indicators (innings, pitches, number of pulls). Backtested on prior seasons, six of the ten bullpens ranked top ten entering September were still top ten through the month, and April-to-August scores correlated with September results at 0.64. A correlation that strong means the right metrics, measured over five months, are worth trusting. It also means about a third of the picture still moves, and usage is a big part of what moves it.

A bullpen report is a short sheet per team: the core relievers graded on skill, their recent trend, and their last three days of work. Built once, it takes minutes a day to update. Before starting, have these in place.

Reading reliever ERA as skill. The tell is a bet justified by a reliever's ERA when he has thrown 30 or 40 innings. Baseball Bet of the Day's point is that 36 innings is roughly 150 batters, which isn't enough to separate talent from noise. The fix is step two: grade on strikeout and walk rates, and use xERA to check whether the ERA is lying.

Backing underdogs because their pen is better. This is the most intuitive angle and it lost money all 2026. Better-bullpen underdogs were 304-369 for -3.2% ROI in mid-September, and 336-405 for -2.8% by Sept. 30, even though Makinen notes the angle had been profitable in past seasons. If you see yourself taking +140 because the dog has the better closer, stop. The bullpen edge didn't overcome whatever made the team a dog.

Expecting the pen to rescue a bad starter. Makinen's own line is blunt: "Not even a strong bullpen can make up for a poor starting pitcher." His data backs it, and it cuts the other way too: fading teams with an elite starter but a negative bullpen rating went 68-59 for -8.6% ROI in 2026. Starter and pen both have to be right.

Chasing two-week heat. The VSiN systems swing hard over short windows. Correlation #3 (elite starters and elite bullpens) was 256-191 and -17.95 units on the season at Aug. 31, yet +9.48 units over the previous two weeks. Correlation #4 went 5-22 in a fortnight. Big favorites of -190 or shorter with a large bullpen edge went 9-0 in the final two weeks, which is nine games. Decide what you bet from a season of evidence, not the last 20 results.

Using a season number for a pen that changed. Arizona's relievers ran a 4.15 ERA through Aug. 29, 2026, then an NL-best 2.68 over the 43⅔ innings after. The Dodgers, presumed contenders, had MLB's worst second-half bullpen walk rate at 12%. Step three exists for this.

Assuming a bullpen game is already priced. It usually is, at least partly. ParlayTools estimates an announced bullpen game moves a posted total by 0.3 to 0.7 runs. If you're betting a total off the news, check whether the number has already moved before you assume you're early.

The fairest case against all of this is that the market is efficient on bullpens and the counting is wasted effort. It's a serious argument, and the broad system's record supports part of it: a 53.3% winner that still lost money is a market pricing quality almost correctly. But almost isn't entirely. The same research found the short-favorite filter up 9.8% on a season, and the postseason fatigue fade below. Quality is priced. Specific uses of it, at specific prices, have not been.

September rosters cap the bullpen. The Sports Room's Sept. 1, 2026 piece lays out the rules: active rosters go to 28, but only two spots are added on Sept. 1 and only one of them can be a pitcher, with a maximum of 14 pitchers. Expanded rosters don't mean fresh arms everywhere. A team that used its pen hard in August doesn't get bailed out. Players also had to be on the 40-man roster by the Aug. 31 midnight deadline to be postseason-eligible, so a September call-up may not be around in October.

Late season rewards the pre-September read. Sawchik's 0.64 correlation is the case for trusting a well-built April-to-August profile in September rather than reacting to a bad week. The same piece notes that the Sports Room links back-to-back appearances above a 1.8 Leverage Index in late August with measurable velocity and induced vertical break decline by mid-September. Watch the overused setup man, not just the team total.

The postseason turns fatigue into the main signal. Makinen's trend: teams playing as underdogs or pick'ems in a postseason series game right after their bullpen threw 5+ innings in the previous game of that series are 41-67 straight up since 2016, for -15.3% ROI. Fading them has a decade of evidence. The rebound angle runs the other way: teams coming off a Division Series game in which their pen blew a save went 35-24 (+17% ROI) in the next game, and 33-23 to the Under.

Wild Card totals lean Under. Wild Card round totals went 40-26 to the Under since 2015 for 20.7% ROI, per VSiN's Sept. 30, 2026 column. Winner-take-all games push the best relievers into more innings. Even here, though, the totals pick OddsIndex made on Phillies at Braves in Game 1 was rated low confidence, because starters weren't confirmed and heavy regular-season workloads could make the late innings volatile. The trend is a lean, not a reason to skip the fatigue count.

First-five bets remove the problem. ParlayTools reports that roughly 56% of runs score in the first five innings. If your read on a starter is strong and your read on both bullpens is murky, the first-five market lets you bet the part you understand.

Early season is mostly noise. The Dodgers' bullpen ran a 4.26 ERA with five blown saves through late April 2026 after signing Edwin Díaz to a record reliever deal, and the Nationals had blown 10 saves across seven relievers early that season. In April, results are a few dozen innings. Lean on strikeout and walk rates, and the prior season's skill, until the sample builds.

The honest summary of the 2026 data is that bullpen work lowers your cost of betting baseball, and in narrow spots turns it positive. Makinen's framing for the broad system's -2.4% ROI is that it still beat the average bettor's -8.1% expected loss by 5.7 points. That's a hold-reduction tool. The profitable angles were the filtered ones: short favorites with a bullpen edge, elite starters paired with elite pens, and postseason fatigue fades. The time cost is real but small once the sheet exists, roughly 10 to 20 minutes a day for a full slate.

One number deserves caution. ParlayTools estimates bullpen ERA rises from about 3.40 fresh to about 4.20 under heavy fatigue, a 0.80-run swing. It's a useful size for thinking about the adjustment, but the source doesn't publish how it was measured, so treat it as an order of magnitude and let your own log tell you whether your fatigue flags are worth that much.