by Keith Law
4
Holtzman’s Folly:
How the Save Rule Has Ruined Baseball
In 2016, Baltimore Orioles reliever Zach Britton had one of the greatest seasons of any reliever in modern baseball history. Britton gave up just 4 earned runs in 67 innings pitched, for a 0.54 ERA on the season, the lowest ever by any reliever who threw at least 40 innings in a season. He gave up only one earned run after April 30. Throwing almost exclusively two-seam fastballs, Britton generated an 80 percent groundball rate, the highest we’ve seen since such data on batted balls first became available. And he was a perfect 47 for 47 in save opportunities.
On October 4, 2016, the Orioles played in the one-game wild-card playoff against the Blue Jays in Toronto, an elimination game for both clubs that determined who would advance to play the Texas Rangers in the next round. The game, somewhat contrary to expectations, was a pitchers’ duel, still tied at 2–2 going into the bottom of the eighth inning, when Britton began warming up in the bullpen. He didn’t enter the game in the eighth, however, as manager Buck Showalter chose to use his setup guy, Brad Brach, in that inning. Britton didn’t appear in the ninth, when Brach started the inning, allowed a double, and later gave way to sidearmer Darren O’Day, who threw one pitch and got an inning-ending double play. Britton—again, who’d had one of the best seasons by a one-inning reliever ever—didn’t appear in the tenth, which belonged to O’Day and Brian Duensing, or the eleventh, when Showalter went to erstwhile starter Ubaldo Jimenez, on the roster as a long reliever. Jimenez allowed a single, a single, and a monstrous home run that ended the game and the Orioles’ season. Britton never threw a pitch.
After the game, Showalter defended his decision to use six relievers in the game but not his best one, saying Jimenez had been one of the team’s best pitchers down the stretch. (True, but still not as good as Britton.) He also cited the concern that he’d use Britton, the Orioles wouldn’t score, and then he’d have to use someone else. Afterward, ESPN’s Dave Schoenfield said of the move, “This is simple: Showalter screwed up.” Yahoo!’s Jeff Passan wrote, “Even the smartest men are capable of ineffable stupidity.” Even O’s fan David Simon, creator of TV series The Wire and Show Me a Hero, got in on the act, quoting his own show with a tweet that read, “Where’s Britton, String? Where’s Britton?”
The truth here is that Showalter was managing to a stat. Because the Orioles were the visiting team, if they took the lead, there would be a save opportunity for Britton, no matter the inning. (Once a game reaches the bottom of the ninth, there can no longer be a save opportunity for the home team.) Showalter mentioned hoping his team would grab the lead so he could get Britton into the game, but that entire line of thinking—don’t use your best reliever to extend the game—is the result of the sport’s obsession with the save rule, which created a statistic that didn’t even exist for the sport’s first hundred years, but a rule that ultimately may have cost the Orioles a playoff victory that night.
A common refrain of the modern dinosaur known as the “sports columnist” is that sabermetrics or advanced statistics are somehow ruining baseball. Jason Whitlock wrote a piece about it, calling analysts “stat geeks,” for Fox Sports in 2011. Someone named Steve Kettmann wrote it in the otherwise progressive New York Times in 2015. What’s even more embarrassing for the Grey Lady is that blogger Murray Chass wrote a piece with the same thesis in 2007 for them, too.
Here’s the truth: no analyst and no sabermetric stat has done half as much to “ruin” baseball as the save, a statistic invented by—you guessed it—a sports columnist.
Jerome Holtzman spent most of his career writing for Chicago newspapers, earning the moniker “the dean of American baseball writers” from his editor at the Chicago Sun-Times, Lewis Grizzard. (Grizzard also remarked in the same essay that Holtzman “had the keys to Cooperstown. No major leaguer ever got into the Hall of Fame if Holtzman didn’t want him there,”* an absolutely horrifying prospect of hubris and power concentrated in one individual.) In an era without any alternative voices—every baseball writer was a white man who wrote for a local newspaper or one of the very small number of national sports publications—Holtzman’s views went unchallenged. This is how we ended up with Holtzman’s Folly: the save rule.
If anyone tried to introduce a statistic like the save today, he’d be laughed all the way to a cornfield in Iowa. The stat is an unholy mess of arbitrary conditions, and doesn’t actually measure anything, let alone what Holtzman seemed to think it measured. Yet the introduction of this statistic led to wholesale changes in how managers handle the final innings of close games and in how general managers build their rosters, all to the detriment of the sport on the field, and perhaps to pitcher health as well.
The idea behind the save was to create an analogue to the pitcher win, but to credit the reliever who recorded the final out or final few outs in his team’s victory. It is true that relievers were undervalued relative to starters at the time, so the impulse to find another way to value their contributions may have been an honorable one if it hadn’t had such a destructive result.
The save rule is the most convoluted of all of the “basic” stats, which only goes to emphasize its arbitrary nature. To be credited with a save under the current version of the rule (MLB Official Rule 10.20), which has been in place since 1975, a pitcher must record the final out in a game that his team won, but one where he didn’t get the win, and the team can’t win by too many runs because then he obviously contributed nothing at all. From the aforementioned rule:
Credit a pitcher with a save when he meets all three of the following conditions:
I. He is the finishing pitcher in a game won by his club; and
II. He is not the winning pitcher; and
III. He qualifies under one of the following conditions:
A. He enters the game with a lead of no more than three runs and pitches for at least one inning; or
B. He enters the game, regardless of the count, with the potential tying run either on base, or at bat, or on deck that is, the potential tying run is either already on base or is one of the first two batsmen he faces; or
C. He pitches effectively for at least three innings. No more than one save may be credited in each game.
I don’t think you could make a more context-dependent stat if you tried. It doesn’t matter how well a pitcher performs in a game if he doesn’t meet all three of these conditions, and he can still meet those conditions while pitching poorly. A pitcher can enter in a save situation, retire every batter he faces, and still fail to earn a save because he allowed an inherited runner—that is, a runner already on base when he entered the game—to score on an out.
The conditions of the save rule create all sorts of internal paradoxes. The pitcher who throws a scoreless seventh inning and a scoreless eighth inning gets nothing, while the pitcher who follows him, throws the ninth inning, and gives up a run but doesn’t surrender the lead gets the SV in the box score. Which pitcher actually contributed more to his team winning the game? Yet which pitcher does Holtzman’s Folly end up rewarding?
The tacked-on bit of rule (c) only adds to the rule’s confusion. If the point of the save was to somehow reward clutch pitching—production in a high-leverage situation that, by definition, only matters if the team wins the game—then why are we throwing on this bit at the end about a pitcher who throws three or more “effective” innings even if his team was up by 15 runs at the time? How does that earn the pitcher the same SV as the pitcher who comes in with the bases loaded in the ninth and strikes out the final batter of the game to preserve a one-run lead? In other works, how does anyone look at baseball and say, “Those two things are equivalent—let’s give them the same stat”?
In 2015, there were 114 relief appearances where the pitcher threw at least three innings and gave up zero runs, but only nine of those appearances earned saves for the pitchers. In eight of those 114 outings the pitcher threw at least 4⅓ innings—recording at least 13 outs�
�and none of those qualified as saves because they all came in losing efforts. Compare those to the thirteen saves recorded by pitchers in 2015 where the pitchers allowed two runs to score in one inning of work . . . but they protected a lead and finished the game, so they got the saves. Four scoreless innings in a loss: no save; two runs in the final inning in a win: save. This is all about the team’s performance, not the pitcher’s, so what exactly are we measuring?
What we think we’re measuring with saves is sort of clear—some sort of intestinal fortitude that marks a man as a Proven Closer™, thus entitling him to millions of dollars in additional salary and to the full confidence of his manager as the Guy Who Pitches the Ninth Inning (Some Restrictions Apply, See Your Rulebook for Details). It’s clear that that’s not even quite true, given the example I just gave above of pitchers earning saves despite allowing two runs in one inning of work. But the save rule also creates a specious separation between closers and nonclosers that doesn’t reflect the relative quality or value of their performances.
On July 7, 2016, Atlanta reliever Mauricio Cabrera did indeed save the day. Atlanta had taken a 4–3 lead over the Chicago Cubs in the top of the eleventh inning, and manager Brian Snitker chose to leave reliever Dario Alvarez, who had pitched a scoreless tenth inning, in the game to try to close it out. Alvarez allowed singles to the first two batters, so the Cubs had the tying and winning runs on base with nobody out. Snitker summoned Cabrera to replace Alvarez, and Cabrera got the first batter to ground into a double play and the second to fly out to preserve the lead. Cabrera was credited with the save and Alvarez with the win.
On August 20, 2016, Arizona reliever Daniel Hudson went one better. With the Diamondbacks leading the San Diego Padres 2–1 in the bottom of the eighth, two other Arizona relievers conspired to load the bases with nobody out, a mess Hudson was asked to come in and clean up without allowing a single run to score. Hudson recorded three pop-ups on seven pitches for a scoreless inning of work, but because he didn’t record the final out of the game, he wasn’t eligible for a save—even though he did what Cabrera did, recording three outs without allowing anyone to reach base or any inherited runners to score, and even had one more runner on base when he entered the game than Cabrera did. The save statistic creates an artificial distinction between these two outings, and the longtime emphasis on save totals has skewed the way we perceive certain relievers based solely on when they’re used.
The most valuable reliever in baseball in 2015 wasn’t a closer and only recorded nine saves on the season. The Yankees’ Dellin Betances, a former top prospect who had trouble just throwing strikes as a starter and moved to the bullpen as a sort of last resort, has recorded some of the best strikeout rates in history since becoming a full-time reliever, and unlike the majority of closers, his usage isn’t limited by the save rule. Betances threw 84 innings in 2015, more than any other MLB pitcher who didn’t start a game, and posted the fourth-highest strikeout rate in baseball. His ERA of 1.50 was the second lowest among all relievers, behind only Kansas City closer Wade Davis, who threw 67 innings, 20 percent less than Betances did. By a statistic I’ll discuss later called Win Probability Added, which measures the impact a player had on his team’s chances of winning each game in which he appeared, Betances had more positive impact on his team’s win total than all but three other relievers in MLB and all but three starters as well. Yet his save total didn’t even reach double digits.
Betances’s usage precluded him from getting saves, but that doesn’t mean he wasn’t used in key situations. Yankees manager Joe Girardi had a Proven Closer™ already in Andrew Miller—also a failed starter who found a second career as a high-strikeout reliever—which allowed him to deploy Betances earlier in games where the outcome was still uncertain. Betances made 74 appearances in 2015, and in 64 of them he entered the game in before the ninth inning began. Of those 74 total appearances, Betances entered the game 16 times with the score tied (thus making him ineligible for a save), and 18 times with the Yankees up by one but where he did not finish the game (also making him ineligible for a save). His low save total doesn’t tell us anything about how Betances pitched, but just when he pitched.
Meanwhile, there were plenty of so-so pitchers who racked up big save totals, something that happens every year because if there’s a close game, there will probably be a save to hand out, and managers today manage to the save rule rather than to win the game. The most egregious example in 2015 was Tampa Bay’s Brad Boxberger, who had 41 saves, the fourth most in baseball, with a 3.79 ERA—right about league average for a reliever.
In fact, there have been seven seasons in MLB history where a pitcher racked up 30 or more saves despite posting an ERA over 5. Two pitched in Colorado and may be excused, since Coors Field is a hitter’s paradise. Brad Lidge of the Phillies did it twice, including a 7.21 ERA in 2009 when he “earned” 31 saves. Joe Borowski led the NL with 45 saves in 2007 despite a 5.07 ERA. These guys did not pitch well, but they were used in a way that got them a lot of saves.
Mike Williams’s 2003 season is one of my favorite examples of save-stat dopiness, where the save column drove the narrative around a player to the point that people forgot to look if the pitcher was otherwise any good. Williams was the token Pirate selected to the 2003 All-Star Game because the Pirates didn’t have a lot of good candidates and Williams had 25 saves at the All-Star Break . . . and a 6.44 ERA, along with more batters walked (22) than strikeouts (19). He was having a terrible year, but the save total fooled someone into putting him on the All-Star Team. A week after the game, the Pirates traded him to the Phillies for a fringe prospect, and Williams didn’t pitch in the majors again after that season.
What the save ends up doing more than anything else is glorifying certain one-inning relief appearances over others. A pitcher throwing a scoreless inning in a major-league game is not that rare; Baseball-Reference’s play index shows more than 1,500 perfect (1-2-3) one-inning appearances by pitchers in the 2015 season alone. There were 469 such appearances that also earned saves for the pitchers in question, but why would we look at those innings any differently than, say, the perfect eighth inning Dellin Betances threw on July 17 of that year, preserving a one-run lead in a game the Yankees would win by that same margin?
(Some of you may see this and think, “Well, that’s why we have the hold statistic for middle relievers.” The hold has numerous problems, including no consistent definition and the possibility that a pitcher can receive credit for a hold without recording an out, but slapping yet another label-stat on a pitcher’s outing merely adds to the confusion rather than resolving it. If saves are stupid, holds aren’t even conscious.)
By giving saves to certain relief appearances based solely on their context in the score, the inning, and the end result, baseball has put a brand on these relievers that, as you’ve probably noticed above, I call the Proven Closer™ even though I’m not actually asserting any legal trademark protection for the term. Much as consumers will pay more for a brand, especially a national brand, than for store-branded products of equivalent quality and that might even be produced in the same facilities as the national brand, major-league teams will pay more for a reliever with saves than an equivalent reliever without saves—in fact, MLB’s own arbitration process ensures that they do so by rewarding saves.
According to the MLBTradeRumors.com Arbitration Projection Model, developed by Matt Swartz, a reliever entering arbitration with 20 saves in his most recent season—called the “platform year” in the vernacular of arbitration—would earn about $1.8 million extra through the arbitration process. Even worse, because subsequent arbitration salaries are based in large part on previous years’ salaries, that onetime raise becomes cumulative, so, according to Swartz’s work, the player would earn at least $4.5 million extra over three years just because of those 20 saves from his first platform season.
The cost wouldn’t be so bad if teams were getting more work out of their closers, but that’s not the c
ase, because managers typically “save” their closers for save situations, declining to use them in other situations that may be important but aren’t save situations according to Holtzman’s arbitrary rule. (Buck Showalter’s decision not to use Britton in the 2016 wild-card playoff typifies this kind of thinking.) In the last five MLB seasons, no reliever who has racked up 30 or more saves has even thrown 80 innings in that same year; only two relievers with 20 saves have done it, and one of those, Jenrry Mejia, only got to 80 innings because he made seven starts before he was moved to the bullpen. It’s not that these relievers can’t handle these workloads; twenty-seven full-time relievers reached 80 innings in a season in that same five-year span. It’s that the save rule has led managers to use their best relievers less often, and that should immediately strike you as both perverse and counterproductive.
It’s easy to just point to the save rule and say that managers are using their relievers to satisfy the rule, but there’s most likely a third variable at work here. If a team gets into a save situation and the manager doesn’t use his capital-C Closer, but then the chosen reliever gives up the lead, the manager will have to answer the postgame question of why he didn’t go to his closer—from the media, from fans, perhaps from his boss. The easiest way to answer these questions is to avoid them entirely by using your closer in save situations and not using him in nonsave situations. It’s suboptimal, but it takes more thought to recognize that burning your closer in a cheap save situation one day may make him unavailable for a tighter situation the next day, and managers don’t face those questions nearly as often.
