How Bigger Bullpens Are Constraining Offense

This is a story about persistence. I thought I had an interesting way of looking at the marked decline in scoring this year so far; as it turns out, I was wrong. A further investigation, however, revealed that another possible culprit was right around the corner. Does it explain the entire decline in offense? Most certainly not. But I’m interested nonetheless, and I hope you will be too.

If you’ve followed baseball in the past five years, you’re probably used to asking questions about league-wide offense early in the season. Major League Baseball has done itself no favors here; the composition of the baseball keeps changing, and home run rates fluctuate wildly as a result. The same is true this year: despite the adoption of a universal DH, offense is down across the board.

The usual suspects are certainly part of the problem. Pitchers keep throwing harder. Putting a humidor in every stadium affects home run rates in unpredictable ways and might suppress home run rates early in the season. The league used two different baseballs last year, and drag coefficient is up this year. Starters are going fewer and fewer innings, giving batters fewer looks at them a third time through the order.

I think that all of those things have something to do with bad offense. But I thought of another potential cause, one I could investigate without learning fluid dynamics. One of the side bargains between the league and the MLBPA after this offseason’s lockout was for expanded rosters early in the season. Teams are allowed to roster 28 players throughout the month of April. On May 30, that number will revert to the standard 26 — this deadline was recently pushed back from May 1. In addition, teams can carry any number of pitchers on their roster until May 1. After that, they’ll be limited to 14, and 13 after May 29.

As you can imagine, teams have taken advantage of these expanded rosters and unlimited pitching allowances to stock up on arms. With starters throwing fewer innings, relievers would normally be heavily taxed, as filling in five innings a day is tough on a bullpen. Teams reacted accordingly; as of Friday, 26 of the 30 teams rostered 15 or more pitchers. There were as many teams with 16 pitchers as with 14. Managers might need multiple pages for a bullpen card these days; every team has an embarrassment of options.

Those extra relievers aren’t just in the bullpen marking time and performing sporadic mop-up duty. They’re out there chipping in, soaking up innings that would otherwise go to starters or gassed relievers. To wit: through Thursday, April 21, there had been 1,406 relief appearances in the 2022 season. Only 11.5% of them, 163 appearances, have been on zero days’ rest — either both sides of a doubleheader or back-to-back games. Last year, 17% of relief appearances were on no rest.

That really matters! Relievers are better when they get to rest between outings. Want a simple proof? I looked at every relief appearance on no rest in the 2021 season. In aggregate, relievers pitching for a second consecutive day produced a 4.63 RA/9. (I chose RA/9 because it’s the cleanest representation of what I’m looking for, namely run prevention.) At first glance, that’s not so bad. That’s quite close to what teams scored per game overall. Hey! Relievers on no rest aren’t any worse than the average pitcher. What no-rest penalty am I talking about?

But that’s not the right comparison. You can’t compare a group of relievers to league average; you’d run into huge selection bias issues. Imagine a world where the only reliever allowed to throw on back-to-back days is Josh Hader. If he produced a 4.63 RA/9 when pitching on his second straight day, that would be a disaster. He’s far better than that, but you wouldn’t notice if you only looked at the aggregate rate statistics.

To account for this, I created a counterfactual: I calculated the RA/9 for every reliever in baseball on days where they weren’t pitching on zero rest. Then I created a control group: the exact same pitchers who pitched on back-to-back days, in the exact same proportion. Instead of their results in those zero-rest games, though, I used their non-zero-rest RA/9 that I calculated up above.

The results? It wasn’t a huge edge, but pitching on zero rest hurt results. Pitchers were worse in pretty much every way. They allowed 0.2 more runs per nine innings. They struck out fewer batters and walked more. They threw slower and allowed more home runs, both overall and per fly ball. In aggregate, those appearances didn’t cost their team much, but they certainly cost them. By having fewer relief appearances on zero days’ rest this year, teams are putting their pitchers in a position to succeed more often. That’s the name of the game; incremental advantages add up to wins.

I could end the story there if I wanted to. The new rules are making bullpens better, which is a huge deal given the massive workload they’ve shouldered so far. The decline in offense? Inevitable. We’re making the good relievers better, and they were already unhittable. Say no more; on May 1, everything will be back to normal.

Only, let’s do the math here. Relievers have thrown 1,585 innings (this is as of last Thursday, when I pulled all the statistics, but the general point holds) and have made 5.5% fewer appearances on back-to-back days; fortuitously, they’re pitching 5.5% fewer innings on zero days’ rest as well. That’s the equivalent of 87 innings. A 0.2 higher RA/9 over 87 innings works out to roughly two runs. That’s not what’s killing offense this year, to say the least.

That doesn’t clear expanded rosters of culpability. Fewer back-to-back appearances might just be a drop in the bucket, but fewer innings by starters is a different matter entirely. Starters have been pretty good this year! They have an aggregate 4.07 ERA, which would be the lowest mark since 2013. They’re throwing harder than ever, and the humidor/ball combination seems to be leading to fewer home runs.

Relievers are pitching a lot better than that. They’re on pace for a season we haven’t seen in decades. Thus far, they’ve posted a 3.34 ERA. They’re striking out nearly a quarter of the batters they face. They almost certainly won’t keep it up; that’s not really how this works, so early in the season. If they did, though, we’d be looking at the lowest relief pitching ERA since 1968.

A lot of this has to do with the ball, no doubt. Relievers strike out so many batters that it’s hard to string a sequence of hits together against them. Sure, they’re prone to walks, but walks hurt most when they’re followed up with home runs. But relievers have always been better on a rate basis, regardless of the baseball. They throw with maximum effort and for fewer innings. It’s no surprise that teams want to use more relievers; pitchers improve when they switch from starting to relief. If you could somehow roster 45 pitchers and throw nine different ones every day, you’d probably put up excellent numbers, assuming you could find enough decent pitching.

Here’s where this “more relievers is always better” mindset might fail. Teams are reaching deeper and deeper into their systems for arms. Eventually, you might surmise, the plan will fail. The relievers you add won’t be as good as the innings your starters no longer have to throw, and the whole calculus will be thrown off.

A counterpoint: we haven’t reached that equilibrium point yet. To test this out, I took preseason Steamer projections for every pitcher who has appeared in relief this year. I removed everyone with no projections — position players pitching, basically — and split the remaining players out into four groups of equal innings pitched, ordered by Steamer-projected ERA. The bottom 25% of innings this year have been pitched by relievers with a preseason projected ERA of 4.98. This is the bottom of the bottom, the guys that teams would prefer to avoid if possible. The top 25% of innings this year have been pitched by relievers with a projected ERA of 3.70 — high-leverage arms and such.

That split isn’t increasing as much as you’d expect. To investigate that further, I chose 2018 as a point of comparison and the equivalent point of the season based on games played, then divided those relievers up into quartiles in the same way. Again, I used their preseason projected Steamer ERAs to assess the talent gap between the best and worst relievers. That year, the high-projection 25% of relievers were projected for an ERA of 4.70; the best 25% of innings went to relievers with a projected ERA of 3.47. The gap was roughly 1.25 runs of ERA for both 2018 and 2022, despite the bloated rosters that characterize this year. In other words, pitching depth is improving so quickly that despite using 25% more relievers this year so far, teams aren’t being forced to sacrifice relative quality.

But wait, there’s more. Here’s those four quartiles in 2022 so far, from high to low in projected ERA. The pitchers who were projected to be worst have done much better than expected, as you can see:

I apologize for using K/9 instead of K%, but that’s the form I had. The point is clear, though: the relievers that projection systems thought would be the worst have been far better than expected. Wondering where your favorite team’s offense has gone? Your team’s best hitter has faced the other team’s worst reliever. Your guy has lost.

Is this a short-lived phenomenon? It’s too soon to tell, but I buy the general direction things are headed in. Pitcher training is much better than it was five years ago. Maybe the very best in the game are doing what they always have, but if you’re a fringe 40-man reliever, the odds that you’re going to Driveline or a similar facility are high and getting higher. The tools to improve are increasingly available; it’s no surprise that pitchers are taking advantage of them.

The 2018 season wasn’t a lifetime ago, but things looked different then:

Is the new normal that relievers in the back of the bullpen will pitch to a better ERA and FIP than starters in aggregate? No, it’s not that either. Some of this is going to come out in the wash. There’s a lot of room between “fringe relievers are improving” and “fringe relievers are putting up sub-4 ERAs”. But if you’re wondering where all the runs have gone, relievers at the end of the bullpen have a lot to do with it — just not in the way I initially assumed.

Statistics in this article are through games of April 21.