Is the Save Finally Slipping?
The first month of the 2017 season hasn’t produced the widespread, Andrew Miller–inspired bullpen changes we thought we might see, but one notable shift has occurred: Saves are being handed out more democratically at this stage of the season than at any point in at least a decade.
Last week, News Tribune beat writer Bob Dutton reported that Mariners manager Scott Servais was planning "a more sabermetric approach" with closer Edwin Díaz, although Servais stressed that his primary motivation wasn’t sabermetrics, but a desire to get Díaz more work. In truth, there’s little distinction between the two rationales: The "sabermetric approach" to bullpen management basically boils down to "Use your best reliever a lot," which the Mariners hadn’t done. Through the team’s first 20 games, Díaz had pitched only six times, accumulating 6.2 innings. In the future, Servais said, "He might go five outs. He might go six outs. In a road game, don’t be shocked if he comes in there in the eighth inning of a tie game. I know it’s not the traditional closer role, but that’s kind of how I see using him."
Since then, Díaz has pitched three times: once in the ninth inning of an 8–0 Seattle win, once in a ninth-inning save situation, and once in a save situation where he entered with (gasp) two outs in the eighth. Maybe Servais will use Díaz in a more adventurous role as the season unspools, but for the most part, it hasn’t happened yet. The 23-year-old righty hasn’t recorded more than four outs or been anything but the last pitcher out of the pen in any nine-inning game in which he’s appeared. That’s the case even though Díaz was one of the best relievers in baseball in 2016, a season that ended with an October (and gave rise to a subsequent spring) in which unorthodox reliever usage became the sport’s cause célèbre.
Here at The Ringer, we’ve been on bullpen high alert, looking for any indication that the fuss over 2016 playoff hero Andrew Miller — which we’ve helped inflame — has inspired other teams to relax their reliever hierarchies and allow longer outings. We’ve conducted preseason and early-season searches for Miller clones, and we’ve talked to the Astros’ Chris Devenski and the Reds’ Michael Lorenzen, two relievers whose usage patterns have strayed from the cookie-cutter norm. But those inquiries have been anecdotal, focused on a few multi-inning arms or late-inning relievers who occasionally enter early.
With April over, we can pull back a bit and investigate whether any bullpen trends reveal innovation on a leaguewide level. There are a few ways we can tackle this: by looking at the length of reliever outings, by checking how closely reliever skill matches reliever leverage, and by examining the distribution of saves.
In an article in March, I ran the following graph, which shows how the length of reliever outings has declined over the past four-plus decades. Both the duration of the average appearance and the percentage of appearances lasting two innings or more have declined dramatically over that span, although they’ve plateaued lately.
Last year’s playoffs trained a spotlight on relievers like Miller and Kenley Jansen, who took advantage of October off days to turn in ironman outings (by 2016 standards) that helped make up for weaknesses elsewhere on their staffs. In the absence of so many off days, though, those outings haven’t become significantly more common, Devenski’s two four-inning performances (and five appearances of six or more outs) notwithstanding. The average innings pitched per reliever game this year is barely up compared with the pre-May rates from the past few years and is lower than 2013’s. (For consistency’s sake, all 2017 stats in this article will be compared with their pre-May equivalents from previous seasons.)
Nor have we seen any notable increases relative to recent years in the percentage of relief outings that have lasted at least four outs or at least six outs:
OK, so relievers aren’t lasting any longer, on average, once they get into games. Are they getting into games more frequently? To find out, I looked up each team’s leader in relief appearances and calculated the average leaguewide percentage of team games in which the appearances leader pitched. The higher the number, the harder-worked the relief leaders were.
Again, no change, at least relative to last year. Which mostly make sense: Not only does a six-month regular season with few days off impose constraints on how hard teams can push pitchers, but valuable long relievers have to be built to break the mold, which can’t happen on a huge scale in one winter. As Devenski told us about his time in the Astros system, "I was kind of developed for this. … I did a lot of different roles in the minor leagues. We had this thing called piggybacking, which was a starter would go five innings, and a reliever, which would be a [regular] starter, would go four innings, and it would be vice versa the next time around."
If managers haven’t been riding their relievers harder, have they at least been deploying them more optimally? Here I looked up the average pre-May "gmLI," or leverage index when entering the game, of every reliever from 2012–17 who had a preseason PECOTA projection. Then I calculated the correlation between their projected ERAs and their pre-May leverage indices. In theory, the lower the pitcher’s projected ERA, the higher his leverage index should be (because his team would want him pitching in the pressure-packed situations that earn higher LIs).
This is kind of a crude measure — it doesn’t account for park factors or the fact that some good pitchers work in lower-leverage roles because they’re ensconced in strong pens — but it does show the fairly consistent inverse correlation that we would expect. (The worse a pitcher is projected to be, the less his manager trusts him to pitch in crucial spots.) And the correlation for 2017 doesn’t look any different from most of the previous years. It’s worth pointing out, though, that all else being equal, leverage index rises as a game goes on, because a team that falls behind by one run in the ninth is less likely to come back than a team that falls behind by one run in, say, the sixth. As a result, a "ninth-inning guy" is likely to have a higher LI than a seventh- or eighth-inning guy, even if he isn’t better. For example, Indians closer Cody Allen has a higher LI than flexible setup man Miller, who hasn’t allowed a run.
It’s not always smart for a manager to make his best reliever a full-time closer, especially if that means using him as sparingly and rigidly as Servais has used Díaz. The ideal is a manager who mixes and matches, determining whom to summon based on availability and matchups instead of assigning one pitcher to every save situation. And that leaves us looking at the distribution of saves.
This time I calculated the percentage of leaguewide saves before May that were recorded by the 30 teams’ saves leaders. The 2017 Mariners, for example, have four saves, all of them earned by Díaz, their saves leader. I identified each team’s Díaz equivalent, combined their saves totals, and then divided the resulting figure by the sum of all saves in the season to date. The results seemed suggestive, so I went back 10 previous seasons instead of five. The finding held up.
Last year’s percentage of saves recorded by save leaders was the highest in the sample. This year’s is the lowest. In the previous 10 seasons, the percentage of pre-May saves recorded by the save leaders averaged 84.9 percent and never fell below 81.6 percent. This year, it dropped under 80, to 79.2 percent.
This could be a fluke, a false positive produced by a random set of circumstances. Last year, Zach Britton recorded 87 percent of the Orioles’ saves; so far this year, sidelined by a forearm strain, he’s recorded only half of the team’s 10. Give Britton nine out of 10 instead of five, and — well, this year would still have the lowest rate in the sample, but not by as much. Throw in Roberto Osuna’s cervical spasm and Jeurys Familia’s season-opening suspension, and it’s easy to explain away what we’re seeing. Of course, closers have gotten hurt or lost their jobs in previous Aprils also, and the percentage still didn’t fall this far.
Granted, the save is nowhere near dead. Almost four out of five saves are still going to one pitcher per team, and 11 teams have let one reliever hog every save. Those 11 include the Diamondbacks, who’ve given 40-year-old Fernando Rodney their highest-leverage bullpen job despite unsettling early results (and Archie Bradley’s strong start in a versatile setup role); the Tigers, meanwhile, remain committed to 35-year-old Francisco Rodríguez as closer despite his own struggles.
Even aside from the injury-impacted Orioles, the teams that have distributed their saves most democratically aren’t exactly led by young bucks in the dugout (although a team’s choice or non-choice of closer is dictated in part by the arms the front office does or doesn’t supply). The Dusty Baker–managed Nationals, who don’t have any one pitcher with more than three of their eight saves, named Blake Treinen their closer out of spring training, hoping to avoid a "by committee" approach. But when Treinen flopped early, Baker replaced him with a Shawn Kelley–Koda Glover duo that became a solo act when Glover went on the 10-day DL with a hip problem. The Angels, meanwhile, have made do with a combo of Bud Norris (five saves) and Cam Bedrosian (three), with one from José Álvarez for good measure. Although Mike Scioscia studiously avoided calling the Angels’ late-inning solution a "committee," a lack of set roles was explicitly part of the plan.
Elsewhere, the Athletics’ Bob Melvin has officially declared a "closer platoon" of lefty Sean Doolittle and righty Santiago Casilla. The Reds’ Bryan Price entered the season determined to loosen set bullpen roles, and he’s largely stuck to the plan. The Phillies’ Pete Mackanin is not naming a closer, after removing Jeanmar Gómez from the role a week after Opening Day.
Some early-season bullpen experiments have short shelf lives; in many cases, situations that seem to be blows to the tyranny of the save can reinforce the old system as soon as one shutdown option emerges. Thus far, though, the save’s stranglehold on the bullpen seems weaker than it has at any point in more than a decade. We may not be seeing the dramatic bullpen upheaval that some daydreamed about last October, but this incremental improvement could be the beginning of baseball blowing the save.
