clock menu more-arrow no yes

Filed under:

The Rays and the story of modern baseball: Abolishing the closer

New, 2 comments

The Kitt-Anderson Debacle, 2015, and Comparative Advantage

New York Yankees v Tampa Bay Rays
Did Boxberger fail so that Karns might succeed?
Photo by Jim McIsaac/Getty Images

Under a system of perfectly free commerce, each country naturally devotes its capital and labour to such employments as are most beneficial to each . . .

By . . . using most efficaciously the peculiar powers bestowed by nature, it distributes labour most effectively and most economically: while, by increasing the general mass of productions, it diffuses general benefit, and binds together by one common tie of interest and intercourse, the universal society of nations throughout the civilized world. It is this principle which determines that wine shall be made in France and Portugal, that corn shall be grown in America and Poland, and that hardware and other goods shall be manufactured in England.

- David Ricardo, On the Principles of Political Economy and Taxation


As he watched Nick Anderson, arguably the best relief pitcher in all of baseball, pitch the ninth inning against the Boston Red Sox with an 8-2 lead, I don’t suspect Kevin Cash was happy.

The day before, August 10, 2019, he had chosen to have Andrew Kittredge come in with one out in the ninth to face the heart of the Red Sox order, protecting a one run league. Kittredge allowed a single to JD Martinez, but then worked around it to earn the save.

Then Cash had Kittredge start what was slated to be a bullpen game the following day, pitching for the fourth time in five days. And Kittredge tore his UCL.

And now, at the end of that following game, here was Anderson protecting a six run lead, because, well, he had to get the work in to stay sharp.

Did the injury happen because of heavy use? Did the quick turnaround add risk? We can’t know that, and even if there is a connection, it’s probabilistic, not causal. All pitching decisions are made balancing competitive needs against levels of arm strain.

But if Anderson had pitched the ninth inning on August 10th, rather than on August 11 (as made the most competitive sense) would there have been less pressure on Kittredge’s arm? And would everyone, Kevin Cash included, have experienced less mental stress?

Absolutely.

Sabermetricians have long criticized the idea of “the closer,” claiming that there’s nothing special about the last three outs of a game, and asserting that pigeonholing your best reliever into only pitching in save situations — most commonly meaning ahead by up to three runs in the ninth inning — is inefficient, because a three run lead with one inning to go is a game that’s already mostly won, and that most pivotal moment in the game often comes earlier. But a one-run lead in the ninth with the best opposing hitters due up is exactly what a closer is for, right?

So what in the world was Cash doing?

Well, it goes back to 2015.


While the 2014 Rays knew exactly how they were going to compete in a tough American League East — on the back of their excellent pitching rotation — the 2015 Rays had a problem to solve. Gone was team ace David Price, traded to the Detroit Tigers. Matt Moore, Price’s heir apparent, would begin the season on the 60 Day IL, rehabbing after Tommy John surgery. Alex Cobb began the season on the IL as well, and would eventually also be destined for Tommy John surgery.

Those were three significant holes to fill, made more difficult when Price’s direct replacement, Drew Smyly, went to the IL in April with shoulder issues that would eventually cause him to miss most of the season.

But then a funny thing happened. The 2015 Rays rotation went out and posted the best ERA in the American League, sixth best in baseball. It’s worth taking a moment to ponder and appreciate this group.

The 2015 Rays Pitching Rotation

Name GS IP ERA FIP xFIP
Name GS IP ERA FIP xFIP
Chris Archer 34 212 3.23 2.90 3.01
Jake Odorizzi 28 169.1 3.35 3.61 3.96
Erasmo Ramirez 27 151.1 3.51 3.76 3.91
Nathan Karns 26 144 3.69 4.05 3.91
Drew Smyly 12 66.2 3.10 3.91 3.47
Alex Colome 13 69 4.70 4.64 4.59
Matt Andriese 8 35.1 3.57 4.15 4.39
Matt Moore 12 63 5.43 4.82 4.81
Steve Geltz 2 4 2.25 2.88 4.36
Data from FanGraphs

How do you replace Price, Cobb, and Moore with Ramirez, Karns, and Andriese and get better? The cliche is that new players aren’t there to replace the stars who are gone, and just need to be the best possible versions of themselves. But the Rays went beyond the cliched psychology of the situation and actually created an environment where those three pitchers could find that best version.

Consider the fact that while none of them were overwhelming starting pitchers, each did something very well:

They were talented but limited pitchers who, if given the chance to focus more on their talents and less on their limitations, might play up in a special way. Which gave the Rays the perfect reason to try something the analysts in the Front Office had probably been itching to do for some time.

The Times Through the Order Penalty

You’ve probably heard more than you wish to about the Times Through the Order Penalty (TTOP) by now. Briefly put, it’s the idea, supported by observation of real-world results, that hitters perform better against an opposing pitcher when they’re seeing that pitcher for successive times in the same game.

This aspect of the game was identified way back in 2006 in The Book, by Tom Tango, Michael Lichtman, and Andrew Dolphin, and it came to real prominence in public domain baseball analysis in 2013 when Lichtman resurfaced the idea, both on his own blog and in a guest piece at Baseball Prospectus.

If you’ve never read and processed those original works, then you should, both because they’re very good, and because they’re Important.

Since then there’s been a lot of good work on quantifying and describing the nature of the effect. I like, for instance, Ethan Moore’s attempts to use his pitch quality metric to further isolate and identify the mechanics behind the TTOP.

There’s also a lot more room for further research.

But if you’re just here for the short version, the things you should know is that:

  • The TTOP means that starters get less effective the more times they face a batter in a game.
  • The TTOP becomes especially pronounced (for complicated reasons that might have to do more about other confounding factors than true effect) the third and fourth times through the order.
  • The TTOP is not new, and is not merely a byproduct of today’s pitchers not knowing how to vary their sequences to pitch late into games. The old pitchers and their fans who claim this? It was happening back then to them, too.
  • The TTOP is not primarily about pitchers tiring. The evidence suggests that a large component of the effect is about hitters adjusting.
  • While the magnitude of the TTOP is not the same for all pitchers (for instance, those with deep repertoires may experience a smaller effect than those with smaller repertoires), it is significant for all. Think of it this way. When a 300 pound wrestler and a 100 pound ballerina each cannonball into a pool from 30 feet up, they’re going to make different size splashes. But if you’re standing next to the pool, the splash from each is going to get you.
  • The fact that a pitcher appears to be pitching well is not a good indication that he will continue to do so, and neither is the reverse — poor performance at the start of a game does not indicate an inability to succeed later on in that game.

2015: The Rays Are Trying Something

So armed with the knowledge that limiting their starters’ exposure to the third and fourth time through the order would likely make those starters more effective on a rate basis, the 2015 Rays did exactly that.

The year previous, a Rays starter had pitched into the seventh inning 67 times, paced by Price at 19 times, Archer at 15, and Cobb at 14. But in 2015, Rays starters only did so 42 times.

Chris Archer, a consistent workhorse at this point in his career, again reached the seventh inning 15 times, but this time he lead the team, and was followed by Odorizzi at 10 times and Ramirez at only 7.

Put another way, in 2014, appearances of at least six and a third innings pitched accounted for a third of the Rays total innings. in 2015, that number was down to only 20%.

Data from Baseball Reference

But 2015 is not a success story. Despite the best starter ERA- in the American League (seven percent above average), and the sixth best wRC+ (exactly average), the 2015 Rays did not make the playoffs. Some of that was bad luck and bad sequencing (they scored fewer runs than their context-neutral batting suggests they should have), but a big part of the story of how those Rays fell out of contention was the bullpen.

Possibly, maybe, the tired bullpen.

The Problem

That 13% of the innings had to go somewhere. Most of it — about 10% of the total innings were still sopped up by the starters, pitching five or six innings, rather than seven, eight, or nine. But that left 3% (just a bit under 50 IP) being picked up entirely by relief pitchers going one inning or less per outing.

Data from Baseball Reference

You can explore an interactive version of this graph here.

On June 30th, the Rays lead the American League East with 42 wins. On July 1st, the Yankees got their 42nd win, and then soon passed the Rays and never looked back.

FanGraphs

Put in terms of FanGraphs playoff odds, the Rays got off to a surprisingly good start, and by the beginning of July had given themselves nearly a 50% chance to make the playoffs. One month later, they were effectively out of the race.

FanGraphs

So what happened in July? Radically oversimplified, the back end of the bullpen — the relievers that those protected starters were turning the game over to early — faltered.

The two most salient examples were Brad Boxberger:

FanGraphs

And Kevin Jepsen:

FanGraphs

Other pitchers followed a similar pattern, with Xavier Cedeño, Brandon Gomes, and Steve Geltz all giving up more runs in July after sustaining a heavy workload through the start of the season.

It’s not just that pulling starters early creates more innings for the bullpen. Pulling starters early while the starters are pitching well puts those innings disproportionately on the top relievers of the bullpen, because most managers choose to call on their best arms when they’re ahead in a close game.

Winning streaks are often hard on the back-end bullpen pitchers anyway, and the 2015 Rays’ early season winning streak and pitching strategy combined to create bursts of work for their best relief pitchers.

Data from Baseball Savant

The 2015 Rays were fifth in the league in “back-to-back appearances,” sixth in “three-in-a-rows,” eighth in four-appearances-in-five-days,” and top third in “three-in-four-days.” You can explore the interactive version of this graph here.

And the load really was falling on the top arms. Boxberger, Jepsen, and Cedeño were all among the top 13 relievers in back-to-back appearance, with Geltz also in the top 50, while Jepsen, Boxberger and Geltz were also 6th, 25th, and 26th respectively in total relief innings pitched (if you’re wondering where Jake McGee is in all of this, he was having his workload more carefully managed while he recovered from offseason elbow surgery).

This graph doesn’t export to an image well, but you can explore individual reliever stacked appearance numbers here.

And so 2015 fell apart, and while the league as a whole was moving toward shorter starter appearances, the Rays went the other way and eased back off the throttle in 2016. Their starting pitchers pitched longer, and their relievers made fewer stacked appearances.

Data from Baseball Savant

2019: The Rays Are Trying Something Again

After two years in the wilderness, and an experimental transition year that provided proof of concept for much of what the Rays were about to do, in 2019 it was time again to commit to unusual workloads and usage patterns.

They dropped the share of innings coming as a part of long start to a level below anything they’d ever done to date, but this time did so while holding down the innings share of short starts as well.

Data from Baseball Reference

It was an extreme shift in innings approach, and it shifted even greater workload pressure onto the relievers than had been applied in 2015. While other teams limited the long starts of their pitchers more than did the Rays, no other team did so while cutting back the short starts as well.

Data from Baseball Reference

(full and interactive version of this source graph here)

If this sounds like a lot of innings for a bullpen to soak up, that’s because it is, and obviously so. It brought the 2019 innings share to 23% for short relief outings, and 19% (the highest in Rays history) for relief outings where a pitcher recorded 4, 5, or 6 outs.

And this is where Anderson and Kittredge come back into the story. The Rays sought to alleviate the impact of the increased workload by cultivating and then actually using reliever depth.

Below is a comparison of the leverage at the time he entered the game for every pitcher to record at least 10 innings in either 2015 or 2019. A leverage index of one means the situation is of average importance to the outcome of the game, with above being more important and below less.

Data from FanGraphs

No 2019 reliever was as highly leveraged as 2015’s Jepsen or Boxberger, and yet six 2019 relievers were more highly leveraged than the fourth highest (Brandon Gomes) in 2015. A full eleven Rays relievers entered games at average or nearly so leverage.

But even the 2019 Rays’ 14th most highly leveraged reliever, Andrew Kittredge, was pitching in a one run save situation when there were better, rested arms available. That’s because the philosophical base of the strategy to spread leveraged appearances throughout a pitching staff is a belief in the Law of Comparative Advantage.

The Law of Comparative Advantage is an idea from the early days of classical economics about how nations should engage with each other in trade. Put simply, and in one of its several forms, it states that if countries produce the good that they are most efficient at producing, and then trades it for the goods they are less efficient at producing, the full system will be more productive and, hypothetically, the people will be happier.

Comparative advantage is a powerful concept. Its effects in the real world have not always been “more happiness,” and it’s questionable whether it even creates the greatest possible long-term economic productivity.

But baseball is a simpler system than worldwide socioeconomics, and comparative advantage has an obvious application. Yes, Anderson was a tougher matchup against right-handed batters than basically anyone else in the league, including Kittredge. He had an absolute advantage. But he was also an incredibly tough matchup against left-handed batters, while Kittredge, a sinker-slider pitcher who did most of his best work dancing down and away off the outside corner to righties, was not someone you wanted facing lefties with the game on the line.

Kittredge held a comparative advantage against righties, and the heart of the Boston order was full of them. If you accept that you can’t call on Anderson every time you want to, and that the Kitts of the team will have to pitch in high leverage at some point during the season, then August 9, 2019, was a perfect time.

In 2019, eleven different Rays pitchers recorded a save falling one short of the MLB record. In the pandemic-shortened 2020, twelve Rays pitchers recorded a save to tie the record. In 2021, they broke the record with fourteen.

The closer was over.

But did it work?

Yes. No. Maybe.

It worked out for Kittredge on August 9th. Then he opened the next day as part of a burst of stacked appearances and he got hurt.

And the Rays really did spread out their leverage and by doing so limit the total workload for their most highly-leveraged relievers. No Rays reliever ranking among the top 30 in innings pitched. To my anecdotal eyes they did not wear down collectively as they had in 2015.

But the graph of stacked appearances is an ugly one for the idea that the 2019 Rays managed to limit starter innings while also protecting aggregate bullpen workload.

Data from Baseball Savant

Exploring this graph on the individual level, there’s some complexity here — the extreme number stacked appearances are coming from heavy use of the high-leverage ROOGY-LOOGY combination of Chaz Roe and Adam Kolarek (aka Rolarek), with each of them often pitching less than a full inning (this was the last season before the coming of the Three-Batter Rule).

But the 31st most used reliever as counted by total innings, Emilio Pagán, also made the fourth most back-to-back appearances in the league, and his appearances were a mix of one and two innings.

All credit to Pagán, who put in an inspiring turn on the wheel when the Rays needed it. But needing it points to a problem with the 2019 approach. Fully buying into comparative advantage in the bullpen is necessary to cut starter innings while protecting relievers from overwork, but it is not sufficient.

There’s a piece that was missing from the 2019 mix, and it’s one the Rays had experimented with in 2018 and then leaned back into after the experience of 2019.

Data from Baseball Reference

In the next part we’ll focus on that missing piece, because the middle-length appearance, once the unglamorous domain of mopup men, is now a job for legitimate bullpen aces.