June 13, 2013
The Truth About Closers and Extra-Inning Games
Over the past week, baseball fans have been treated to some epic extended baseball games. This past Saturday, the Rangers and Blue Jays played 18 innings before they could decide the matter, and that wasn’t the longest game that day. The Mets and Marlins played 20 innings in their game. It made last Wednesday's White Sox-Mariners 16-inning game look like a quickie.
That flurry of bonus baseball reignited a conversation about bullpen management within extra-inning games, specifically for the visiting team. In the Dallas Morning News, columnist Evan Grant talked about the issue in relation to the Rangers-Blue Jays game, which was played in whatever they're calling the SkyDome now (I know, the Rogers Centre). In that game, Rangers manager Ron Washington found himself a little short in the bullpen with two relievers (Tanner Scheppers and Michael Kirkman) not available to pitch. Going into the bottom of the 12th inning, with the game still tied, he was down to closer Joe Nathan and long reliever Ross Wolf. He sent Wolf out to face the Blue Jays and let Wolf throw 82 pitches over 6 2/3 innings before the Blue Jays finally pushed across the winning run. Nathan, arguably the best reliever Washington had, and certainly better than Wolf, never left the bullpen. Should he have?
There is a strange unwritten rule in baseball that a visiting manager should not use his closer in an extra-inning game until his team gets a lead. After all, it's not a save situation. I've always found this to be a silly practice. After all, in the bottom of the ninth inning (or an extra inning) of a tie game, if the home team scores, the game is over. There is no margin for error. In a save situation, the team has the lead. If the pitcher concedes a run, you live to fight another inning. Why have the inferior pitcher throw when there's no safety net and the superior pitcher throw when there is one?
I also find it perplexing that when I bring up this particular strategic blind spot, even when people understand why it's a blunder, they still ask that annoying question, "But who will get the save?"
Part of it is that we've reached the point in human history where the word "closer" has come to mean "he who accumulates saves." It's unthinkable that anyone other than the anointed one would have the letter "S" after his name in a box score. But there's another real human weakness behind why people would rather commit a strategic blunder and save the saver for the save situation. It's called loss aversion.
People (often irrationally) overvalue what they have and react more negatively when they lose something rather than simply not gaining it in the first place. There's an emotional reaction to blowing a save that isn't the same as just having the other team break the tie. Managers (and fans) seem to want their best reliever on the mound when the emotional stakes are the highest rather than when it would strategically make the most sense.
Or would it?
A road team that doesn’t score in the top of an extra inning is going to need, at minimum, two more innings from its bullpen if it wants to win the game. The Rangers needed someone to shut the Blue Jays down in the bottom of the 12th, to score in the top of the 13th, and then hold the line in the bottom of the 13th. And that's if everything went right. From the sound of it, there were only two options available to Ron Washington (Nathan and Wolf). Perhaps Nathan could have been stretched to two innings—he was once last season—but if the Rangers didn't score in the top of the 13th, Wolf would have had to pitch at some point anyway, possibly even in a—gasp—save situation.
Perhaps we should instead look at the situation for what it really was. Ron Washington had two pitchers left. One of them was a good guy who could go one inning, and the other was a guy who could throw a lot but wasn't as good. In other words, he was in a bad spot. There just isn't a way to spin that. He could have pitched Nathan in the 12th, and then he probably would have had to go to Wolf in the 13th with no one left in the pen (unless he wanted to use starters on their "throw" days). He could have pitched Wolf until he got a lead and then used Nathan, who might have coughed up the lead himself and then be stranded out there. (The White Sox had this problem in their 16-inning game: closer Addison Reed was the last man standing in their pen and ended up pitching three innings after the Mariners came back to tie the game in the 14th.) No matter what he tried, there was a high likelihood that it wouldn't work. Think managing a major league team is easy? Tell 'em, Wash. It's incredibly hard. But no matter how bad the options are, there is always one that is better—or perhaps less bad—and that's the one you should choose.
Joe Nathan should have pitched the 12th inning. In fact, Nathan probably should have come into the game earlier. When there is no margin for error (and starting in the bottom of the ninth, if the game is tied, there is no margin for error), you want your best available pitcher on the mound. When he has to leave, the next-best pitcher should take over.
There are a couple of objections to this strategy. Evan Grant mentions that relievers often prefer set roles. (A manager could define this as part of a reliever's role, no? Ah, yes, but his agent prefers a defined role that involves him racking up a lot of saves.) And perhaps pitchers perform worse when taken out of their role of protecting leads and charged with protecting a tie.
And further, sending Ross Wolf out for a save situation probably increases the chances of one of those heartbreaking losses. While they all count the same in the standings, the heartbreak might linger to the next day and affect the team. If you're going to lose, why not lose in a way that minimizes heartbreak? (Why on earth are you managing like that?)
Warning! Gory Mathematical Details Ahead!
Per usual, I used the log-odds method to estimate and control the likelihood of various events, given the batter-pitcher matchup at hand. Pitchers in tie-game situations were actually less likely to strike hitters out as well as less likely to give up a home run. The probability of an out in play went up. And that makes sense: in a tie game, a pitcher would tailor his gameplan away from things that might cause a ball to fly over the fence, which would cause his team to lose. In a save situation, he has the lead, by definition, and while coughing up a lead is bad, it's not fatal.
There is evidence to suggest that relievers do pitch differently when protecting a tie vs. protecting a win. However, the differences pretty much washed out. They aren't really better or worse off, and as an added bonus, closers in general pitch in a way that fits the situation.
Then there's the question of whether not having the closer to protect a lead might lead to a particularly heartbreaking loss. I found all games in which a team had gone into the ninth inning or later with a lead, but surrendered that lead and lost the game. I then looked at what happened in the next game for that team for both their pitchers and their hitters. Again I used the log-odds ratio to control for batter and pitcher matchups. There was no evidence that a heartbreak the night before predicted any deviation from what would have been otherwise expected. I expanded the framework to whether a meltdown had happened in any of the past five games for the team. Still no significant effects.
Teams may feel awful about what happened last night, but it doesn't seem to affect their performance tonight. And I would argue that this shouldn't surprise anyone. In the immediate aftermath of a loss, it's tempting to believe that all hope is lost. But let's remember that a baseball clubhouse is not a static entity. Time doesn't stop after the winning run crosses home plate. Baseball teams have their own culture that helps them to deal with the realities of baseball in general, and all cultures have ways to help members deal with loss.
Consider that every culture has some formal ritual for dealing with the ultimate loss, death. A group of baseball players will have a lot of incentive to figure out a way to deal with losing games, and they will have their rituals for doing so. Joe Maddon of the Rays has a 30-minute policy for both wins and losses. The team may celebrate (or mourn) for half an hour, and then must move on. The ritual doesn't have to be that formal, but it exists in every clubhouse in MLB, and if it doesn't, it's a problem that needs to be fixed.
So when should I bring in the closer?
Let's return to the case of Ron Washington's decision between Joe Nathan and Ross Wolf. About the only other advantage to holding back Nathan from the game was that Nathan was then fresh for the next day (in fairness, when Texas would have a thin bullpen). Then again, why have a pitcher you pay $7 million a year if he doesn't pitch in the close games when everything is on the line? Tomorrow, it may rain.
Wolf did manage to pitch six shutout innings before Toronto got to him in his seventh (not bad). Sometimes you make the wrong decision and it works out anyway. It doesn't change the fact that it was a bad decision. Nathan may have come into the 12th and given up a home run to the first hitter he faced. It happens.
But the reluctance to use a closer on the road in a tied extra-inning game has to stop. Inferior relievers are being put into situations that call for the best available option, all to serve a human failing (loss aversion) and the weird idea that all a "closer" is allowed to do is pitch in save situations. Managers (and fans) are making decisions based on their emotional state, and this is a time that calls for cold, calculating rationality. That's not always comfortable, but there's a difference between the strategy that feels the best and the one that is most likely to win you a ballgame.