Skip to main content
Menu
Relievers Have Broken Baseball. We Have A Plan To Fix It.

Earlier this month, Major League Baseball said it was considering a rule change to require pitchers to face at least three batters per appearance — or finish an inning — as part of a series of initiatives to improve the pace of play. I don’t hate this; I’ve always been a fan of relief pitchers working longer outings. But I think the MLB proposal misses the real problem.

The issue isn’t really with relievers who face just one hitter at a time. In fact, LOOGYs — Left-handed One-Out Guys — are already fading in popularity as teams realize that if a pitcher isn’t good enough to face multiple hitters in a row, he may not belong in the bullpen pecking order at all.

Instead, the problem concerns teams that use a parade of relievers who enter the game from the sixth inning onward and throw the hell out of the ball, knowing they’ll probably max out at one inning at a time. (The Yankee bullpen is a prime example.) You might call these pitchers OMGs: One-inning Max-effort Guys. They can be incredibly, game-changingly effective, but they aren’t necessarily all that skilled.

In fact, the whole problem is that OMGs are a renewable resource, with no real constraints on supply. Teams can take failed starters with two decent pitches and, after some weeding out, turn them into OMGs who will strike out 25 or 30 percent of the batters they face, provided they only have to throw one inning every second or third day. It also yields rosters that are grossly imbalanced relative to the amount of value that these relievers generate. According to FanGraphs, relief pitchers accounted for only about 9 percent of the value (in wins above replacement) that all position players and pitchers created last year. And yet, they occupy about 25 percent of roster slots.

And to a larger degree than you probably realize, these OMGs bear responsibility for the ever-increasing rate of strikeouts in baseball — something that was easier to shrug off until MLB attendance started to decline.

More relievers means more strikeouts

Strikeouts have been increasing for more or less the entirety of baseball history. Here’s the trajectory from 19081 up until last year — when, for the first time, more plate appearances ended with strikeouts than with base hits.

There are a couple of peaks marking the end of the Deadball Era in the late 1910s and then another pitchers’ era in the mid-to-late 1960s, but overall the trend is very steady. Over this period, the correlation between the year and the strikeout rate is 0.91.

One other baseball trend has been equally if not more relentless, however: As time has passed, teams have relied more and more on their bullpens. As a result, both starting pitchers and relievers have seen increasingly shorter stints. Thus, the number of pitchers per team per game has steadily increased, from 1.4 in 1908 to around 4.4 now.

The correlation is stronger still if you look at the number of pitchers used relative to the number of plate appearances in a typical game.2 For instance, if you take the number of pitchers used per 38 plate appearances3 — over the long run, MLB teams average about 38 plate appearances per game — you get this:

That looks a lot like the previous graph showing the strikeout rate — the correlation is 0.96 — including a dip in both pitchers used and strikeouts at the end of the Deadball Era in the late 1910s and again at the end of the Second Deadball Era in the early 1970s, and then an especially steep acceleration in both strikeouts and pitchers used over the past few years.

It’s not just a coincidence that relief pitcher usage and strikeout rate are correlated in this way. When you take a starter and use him in relief — especially in a short stint that typically lasts only an inning or so — his strikeout rate will be usually be higher, and sometimes a lot higher. You can also expect him to throw harder and to use a more dangerous repertoire consisting of more fastballs and sliders.

Here’s the tale of the tape. Using data from FanGraphs, I looked at all pitchers who worked both as starters and relievers between 2016 and 2018, providing for a direct, head-to-head comparison of how the pitchers performed in each role. These pitchers’ strikeout rates were about 12 percent higher when they came on in relief than when they started. They also threw about a mile per hour harder in relief.4

Starters supercharge their K rate when working in relief

Statistics for MLB pitchers who worked as both starters and relievers, 2016-18

As starter As reliever
Strikeout rate 18.4% 20.6%
Fastball velocity 91.6 mph 92.5 mph
Share fastballs 54.1% 55.1%
Share sliders 13.9% 15.0%

Observations are weighted by the lesser of the number of batters a pitcher faced as a starter and in relief from 2016 to 2018. For example, a pitcher who threw to 500 batters as a starter and 200 batters as a reliever would receive a weight of 200. Pitchers who averaged fewer than 15 batters faced per start, i.e. who served as “openers” or tandem starters, are excluded from the analysis.

Source: Fangraphs

Those are meaningful gains, but the really big differences come when you use pitchers in short stints that are roughly one inning long. In the next table, I’ve assigned the pitchers who worked both as starters and relievers into three groups: first, those who averaged five or fewer batters faced per relief appearance (these are guys who usually threw just one inning at a time — the OMGs); second, those who averaged more than five but fewer than eight batters faced (a mix of one-inning and multi-inning appearances); and third, those who averaged eight or more batters faced (mostly multi-inning appearances).

It’s much easier to throw an inning at a time

Statistics for MLB pitchers who worked as both starters and relievers, 2016-18, by how many batters faced per relief appearance

Five or fewer batters
As starter As reliever
Strikeout rate 19.9% 23.9%
Fastball velocity 91.7 mph 93.6 mph
Share fastballs 53.6% 56.9%
Share sliders 17.7% 19.4%
Between five and eight batters
As starter As reliever
Strikeout rate 18.7% 20.6%
Fastball velocity 91.5 mph 92.3 mph
Share fastballs 53.6% 54.0%
Share sliders 12.6% 13.6%
Eight or more batters
As starter As reliever
Strikeout rate 16.7% 17.7%
Fastball velocity 91.6 mph 92.2 mph
Share fastballs 55.6% 55.8%
Share sliders 13.4% 13.9%

Observations are weighted by the lesser of the number of batters a pitcher faced as a starter and in relief from 2016 to 2018. For example, a pitcher who threw to 500 batters as a starter and 200 batters as a reliever would receive a weight of 200. Pitchers who averaged fewer than 15 batters faced per start, i.e. who served as “openers” or tandem starters, are excluded from the analysis.

Source: Fangraphs

The first group — the OMGs — got a massive, 20 percent boost to their strikeout rate as relievers. They also gained about 2 mph worth of fastball velocity. And they were able to throw fastballs or sliders — the pitches that seem to be at the core of increasing K rates — 76 percent of the time in relief as compared with 71 percent of the time as starters.

Conversely, the third group — the long relievers who routinely worked multi-inning stints — got only a 6 percent gain in their strikeout rates relative to the ones they had as starters, and they added only 0.6 mph to their fastballs.

LOOGYs aren’t really the problem

The MLB proposal would effectively kill off the LOOGY, along with its much rarer companion, the ROOGY. So it’s worth asking: If relief pitchers are especially effective when they’re limited to only one inning of work, does it follow that they do even better when limited to just one or two hitters? That is to say, could MLB’s proposal to require that pitchers face at least three batters cause an especially large reduction in strikeout rates?

The answer is: not really. If you further break down our sample of pitchers and look at those who threw very short stints in relief,5 they actually had fewer strikeouts than those who averaged around an inning per appearance.6 A lot of this is selection bias: Guys who are brought in to face only one or two hitters at a time are usually mediocre pitchers with big platoon splits. Left-handers who became LOOGYs are generally worse as starting pitchers than the rest of the sample; indeed, they’re quite a bit better in relief than in their starting roles. Nonetheless, they’re not all that effective in relief — much less effective than the OMGs — and because they throw so few innings, they don’t affect the bottom line that much in terms of baseball’s strikeout rate.

And because LOOGYs are fading in popularity, they don’t necessarily contribute all that much to slowing down the game. Of the roughly 16,000 pitching changes in 2018, only about 5,000 occured in the middle of the inning, according to data provided to FiveThirtyEight by David Smith of Retrosheet. These midinning changes are indeed time-consuming — adding about 3 minutes and 15 seconds worth of game time, Smith estimates. (Pitching changes between innings add only about 15 seconds, by contrast.) But they aren’t all that common.

How to bring balance back to bullpens

There’s a better idea than the MLB minimum batters proposal, one that would also speed up the game but that would yield more interesting strategy and — most importantly, from my point of view — cut down on the number of strikeouts, perhaps substantially. The core of my proposal is simple: Each team should be limited to carrying 10 pitchers on its 25-man active roster, plus an Emergency Pitcher.

Like it? Hate it? Well, let me give you some of the details first:

  • What’s an Emergency Pitcher? He’s a pitcher who could be signed either on a game-by-game basis — in the way that emergency goalies are used in the NHL — or for any length of time up to a full season. The Emergency Pitcher couldn’t be a member of a team’s 40-man roster, although — just for fun — he could be a member of a team’s coaching staff.7 Emergency Pitchers could enter the game only under certain circumstances:
    1. If the starting pitcher left the game because of injury;
    2. If one team led by at least 10 runs;
    3. If it were the 11th inning or later; or
    4. If it were the second game of a doubleheader.
  • Position players could still pitch, but they wouldn’t be allowed to pitch to a greater number of batters than the number of plate appearances they’d recorded so far on the season as hitters. A backup catcher with 100 plate appearances could face up to 100 batters as a pitcher, for instance (which works out to roughly 20 or 25 innings). With this rule, teams could use position players to pitch on an emergency basis basically whenever they wanted, but they couldn’t designate pitchers as position players just to circumvent the 10-pitcher requirement. Brooks Kieschnick types would need to have their innings and plate appearances monitored carefully.8
  • After the roster expanded to 40 players in September, minor league call-ups who were not on the 10-pitcher list could start games, subject to a requirement that they threw at least 60 pitches or five innings or — a mercy rule — gave up at least five runs. They could not appear in relief, however.
  • Relief pitchers, especially the OMGs, aren’t going to like this, so the restrictions could be phased in over several years. For instance, you could start with a 12-pitcher limit beginning in 2020, then ratchet it down to 11 pitchers in 2022 and 10 pitchers in 2024 as teams adapted to the new requirements.

As you can see, the goal here is to be fairly strict: While we want to provide for a bit of flexibility, we mostly want to force teams to stick to the 10 players they designate as pitchers as much as possible. For that matter, we’d probably also want to tighten rules surrounding the injured list and minor-league call-ups, which teams regularly use and abuse to add de facto roster slots — but that’s not a part of this proposal per se.

Toward a new equilibrium

So how would teams use their pitching staffs under these rules? That’s anyone’s guess, and part of the fun would be in seeing the different strategies that teams adopted. But my guess is that the average team would do something like this to fill the roughly 1,450 innings that major league teams pitch in each regular season:

What a 10-man pitching staff might look like
Role Games Pitched Games Started Innings Pitched
Ace starter 34 34 230
No. 2 starter 33 33 210
No. 3 starter 33 33 195
No. 4 starter 32 32 180
No. 5 starter 30 22 150
Long reliever/spot starter 40 3 100
Durable middle reliever 55 0 90
RH set-up 60 0 85
LH set-up 70 0 75
Closer 60 0 80
Role Games Pitched Games Started Innings Pitched
Emergency Pitchers 10 0 20
September call-up starters 5 5 25
Position players 5 0 10
Games Pitched Games Started Innings Pitched
Total 467 162 1,450

This strategy envisions that starting pitchers would throw 6.0 innings per start, up from 5.4 innings per start in 2018 but a bit less than the 6.2 innings per start that pitchers averaged in the 1980s. Relievers would average around 1.6 innings per appearance, meanwhile — considerably up from 2018 (1.1 inning per appearance) and about the same as in the 1980s.

Overall, this plan would entail using 2.9 pitchers per team per game, which is close to where baseball was in the late 1980s. But we could balance out the workload more effectively than teams did back then. As you can see in the table, we could get the necessary innings from a 10-man staff without having to ask starters to throw 270 or 280 innings, as ace starters sometimes did in the 1980s, and without having to ask closers to throw 140 innings a year, as sometimes happened too. Starters would have to work through the third time in the order a bit more often, but there would still be plenty of room for discretion on the part of the manager.

The most consequential change would be that we’d cut down on the number of OMG innings. There would still be plenty of them, to be sure. But if you went overboard, it would come with a lot of trade-offs. If a team tried to employ five relievers who each worked 70 appearances of one inning each, for instance, its five starters would have to average about 6.5 innings per start, so they’d be working through the third time in the lineup a lot more often.

And if you did want to use a pitcher to face only one or two batters, you could still do it, but it would be more costly still — with a 10-man pitching staff, someone else is always going to have to pick up the slack.

This would also relieve (pun somewhat intended) the monotony of the OMGs. We wouldn’t be removing any spots from the 25-man roster. (In fact, we’d essentially be adding one for the Emergency Pitcher.) But we’d be requiring at least 15 of them to be used on position players. Pinch runners, pinch hitters, platoon players, defensive replacements and third catchers — all of whom have become endangered species as teams use every marginal roster slot on an OMG — would begin to roam the baseball field freely again.

I’m reluctant to estimate the overall amount by which my rule change would reduce strikeouts or improve pace of play. That’s because baseball strategy is a dynamic system, and our goal is to change teams’ overall attitudes toward pitcher usage. Pitching to contact might become more common again, for instance, as starters would need to throw longer outings. Keep in mind that if starters are only expected to work through the order two or two-and-a-half times, tossing perhaps five or six innings, they can also throw at relatively high effort. So we wouldn’t just be reducing strikeouts by exchanging some OMGs for multi-inning relievers; starters would also have to pace themselves more, too.

But if relief-pitcher usage has as close a relationship with strikeout rates as I think it does, the net effects could be substantial. This rule would essentially roll relief-pitcher usage back to what it was in the late 1980s or early 1990s and could bring strikeouts back toward what they were back then too, when pitchers struck out about 15 percent of the batters they faced instead of the 22 percent they do now.

That’s probably too optimistic; at least some of the increase in strikeout rate undoubtedly has to do with pitchers being bigger and stronger and throwing harder than ever before.9 But some kind of intervention is needed. The OMG-dominated equilibrium of today may be ruthlessly efficient, but it isn’t making for an aesthetically or strategically rewarding form of baseball.

Footnotes

  1. I’m using 1908 as the cutoff because that’s the earliest season for which Baseball-Reference.com has data on the number of pitchers used per game, which we’re comparing the strikeout rate against.

  2. This accounts for the fact that other things held equal, strikeouts reduce offensive output, and less offense means fewer plate appearances per game, since the team doesn’t get through the order as often.

  3. More precisely, per 38.23 plate appearances.

  4. In my analysis, observations are weighted by the lesser of the number of batters a pitcher faced as a starter or as a reliever. For example, a pitcher who threw to 500 batters as a starter and 200 batters as a reliever would receive a weight of 200. Pitchers who averaged fewer than 15 batters faced per start, i.e. who served as “openers” or tandem starters, are excluded from the analysis.

  5. Those who averaged fewer than four batters faced per relief appearance between 2016 and 2018.

  6. Four or five batters faced.

  7. Maybe Bartolo Colon could play into his 60s as an Emergency Pitcher/pitching coach.

  8. Or teams could designate their Kieschnicks as pitchers; nothing in what I’m proposing would prevent a team’s 10 pitchers from being used at other positions.

  9. Then again, hitters are probably also better than ever before.

Nate Silver is the founder and editor in chief of FiveThirtyEight.

Comments