Baseball Prospectus is looking for a Public Data Services Director. Read the description here.

There’s nothing more thrilling in baseball than a ninth-inning comeback. Unless, of course, it’s your team being victimized by the comeback. Then, there’s nothing worse. To have fought for eight innings and held the lead, only to have the game snatched away in the ninth. It might leave the other team breathless, but it will leave you with a nasty scar.

Of course, this is one piece of how the modern “closer” developed to be “the guy who protects small leads in the ninth inning.” Even if it’s not the mathematically optimal use of the closer, there’s a certain logic to saving the closer for when the psychological pain of coughing up the lead is the greatest. There’s going to be a game tomorrow and you don’t want players moping around tomorrow night thinking about last night’s misfortunes. There’s the standard counter-argument that if you let the game get away in the eighth, it counts the same in the loss column as if you let it get away in the ninth, but that doesn’t seem to work. It’s that finality piece that they come back to. If the eighth and ninth are left and the set-up guy and closer are both going to have to pitch, then it’s better to slowly choke off the other team’s chances by deploying them in the “correct” order. Or at least it feels better to do it that way. And besides, the memory of this game will linger long into the echoes of eternity or something like that.

In this way, the argument for saving the closer/best reliever for the ninth inning makes a certain amount of sense. Using the simple framework of two relievers, one the “setup guy” who usually pitches in the eighth inning and the other being the “closer” who pitches the ninth inning, it might not be the optimal way to use those guys, specifically in this game, but if we assume that losing in the ninth hurts in a special kind of way, one that will make future games more difficult to win, suddenly, it becomes a math equation that looks at variables like “How much more difficult?” and “How long does the effect last?” If the expected cost is more than the expected gain in efficiency by using a non-“traditional” model (insomuch as the ninth-inning-only closer is “traditional”), then it’s actually worth it to stick to the hard line of the closer pitches the ninth.

The thing is that while I accept that such a carryover-feelings boo-boo could actually be real, the two most important words in science are “prove it.”

Warning! Gory Mathematical Details Ahead!

Let’s look at three types of games. Games that were lost in the seventh inning (the pitching team enters the pitching part of the seventh inning with a lead, but ends up losing the game), games lost in the eighth inning (I think you can figure it out), and games lost in the ninth. It’s possible that a team could lose a lead in the seventh, storm back and re-take the lead in the eighth, only to lose it again, and do it all over again in the ninth. In cases where the team blew multiple leads, I assigned the game to the latest applicable inning where a lead had been lost and the game was eventually lost.

I suppose if there’s going to be an effect, it should be the strongest the next day. For those teams, I looked at what happened the next time the team played a game. Data are from 2000-2015.

Lost Lead in

Winning Percentage in Next Game

Expected Win Percentage Based on Seasonal













We see that after a hard-luck loss, teams are less likely to win the next day. But look at the numbers by inning. The “penalty” is actually most harsh when the game was lost in the seventh. For what it’s worth, the take home message here probably isn’t entirely that losing a game in the seventh sends teams into a tail-spin. (Teams that we otherwise would expect to be on an 80-win pace suddenly become teams on a 71-win pace!) Losing a game in the seventh or eighth or even ninth might actually be a signal that this was the time of year in which a team didn’t have all of its good relievers healthy, and the next day they are still operating without those key pieces. It’s also possible that we’re just picking up on the fact that a team that’s likely to pull off a big comeback is just a good team in general, and our victimized team is more likely than not to still be playing them the next day. Still, I could see there being some effect of blowing a lead, but the data actually point away from the ninth inning being the danger zone. So, it’s bad to blow a lead in general, but there’s nothing magic about the ninth inning.

To go a little deeper, I looked at the at-bat level (this time with data from 2011-2015). I looked at all hitters either the day after their team had blown a lead (in the 7th, 8th, 9th , in general) vs. no such catastrophe. I controlled for batter and pitcher quality using the log odds ratio method. Over a range of outcomes, in general, losing a lead in the 7th (or 8th or 9th or in general) the night before had hardly any effect on how batters performed the next day. I say in general, because I looked at four different situations (lost in 7th, 8th, 9th, or in any of them), and over several different outcomes, and found a few isolated significant coefficients, but this is likely the building of Type I error. Which outcomes turned up significant were not consistent, and the findings that were there were mixed as to whether they were good or bad for the batting team.

I then turned the model around to look at how pitchers performed after their team blew a game the night before, again controlling at the at-bat level for batter and pitcher quality. In general, there appeared to be no consistently significant effects. Well, if there’s no effect on the next day, then when exactly will that effect pop up? Considering the fact that players the next day tend to do about as well as we would expect (both on the pitching and hitting side), but they do still lose more, the data support the hypothesis that this is more a case that either they are not fielding their best team or they are more likely to be playing a good team.

Maybe it’s not just one game that does the trick. One game can be a fluke. You’re angry that night, but you realize that this is part of baseball and it does no good to be angry. But what happens when it’s happened five times this year? Does that take its toll over time?

I created an index of how often a team has been “burned” over the course of the season. I did it “to date” so that if at the start of game 120, the team has lost three games in the ninth, the regression knows that. If they do so again that night, for game 121, that number has escalated to four. To make sure that we’ve given the team the chance to suffer the slings and arrows of outrageous fortune, I looked at how they performed from game 100 onward.

I created a logistic regression for each game from game 100 to game 162, again controlling for the team’s overall seasonal win percentage, and then an indicator of how many times, to date, they had been victimized by a comeback in the (7th, 8th, 9th, in general). It turned out that the number of comebacks suffered in general was the most predictive variable, but that the variance that this explained was more than 100 times less than just knowing the team’s win percentage. When I ran similar analyses at the plate appearance like I did above, this time using the cumulative number of comebacks a team had suffered, I found little evidence that it makes a difference there either.

If there is a hangover effect from being the victim of a comeback, it is very weak and it seems that what effect there might be is not actually dependent on the inning in which the comeback took place. A late loss is a late loss is a late loss. What that means is that teams should not worry about preventing a ninth-inning comeback, but instead, find the best way to prevent any sort of comeback. Not only is that the point of tonight’s game, but also because comebacks in general seem to be (as much as there is an effect) the driver of any deleterious effects in the future. If it makes sense to deploy the closer earlier in the game because that’s when the game is really on the line, the counter-argument that it might leave a team more vulnerable to a ninth-inning comeback (and that’s what you really need to look out for) is not supported by the available evidence.

The Summer of the Non-Closer
While we didn’t see much of a hangover effect for a team that had been victimized by a comeback, either in the short- or long-term future, there certainly is a belief in general that one exists. Where does it go? For one, I think baseball players are a little more resilient than we give them credit for. No, it isn’t fun to lose a game that you thought you were going to win, but that happens. People deal with much greater adversity on a daily basis. But within the clubhouse, there is likely already a ritual to put losses behind them. They can scream expletives or something like that. It’s a natural feature of humans. We have ways of coping with unfortunate events. We have funerals.

If there’s one possible worm-out available, it’s that in most of the ninth-inning comebacks that teams have had to endure, they at least used the now-standard “closer in the ninth” model, meaning that they could frame the loss as “Well, we did everything we were supposed to do and it didn’t work… that happens sometimes.” If a team were to start using a non-standard closer model and lose a game in the ninth, it might be seen differently and maybe that would have a different reaction.

Still, the idea that “We need the closer because the last three outs are the hardest to get and we don’t want the guys sulking” prevents us from using a slightly more efficient bullpen model is actually not a good argument, at least according to what we know from these analyses. So managers, feel free to use your best pitcher—dare I say your closer—whenever he is most needed, even if that’s not the ninth inning. Instead, work to make sure that when a comeback does happen (and it will at some point) you have a good way to help the guys deal with it.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
Nice analysis
Thought provoking article. Sunday nights game, Red Sox vs. Astros was a classic study. Red Sox lead by 2, 2 outs in the 9th with Kimbrel pitching, and disaster strikes. As a dyed-in-the-wool Sox fan I was devastated, well at least shookup severely, but now looking at the other side of the coin, could the fact that the Sox came again and won while getting 3 innings from a back end of the bullpen pitcher be a very positive thing. The Sox have followed up that fortuitous win with 2 more, all be it against a Triple-A team.
Pizza: I think if you split it up between home/road, then the effect is reduced substantially.

So, the team that loses the lead is most likely to be: (a) below average, as your last column shows, but also (b) be the road team.
Hadn't thought of that angle. Makes sense. Personally, I'm more interested in the non-differences between the 7th, 8th, and 9th inning comebacks. My original title was "Tis Better to Have Loved and Lost... in the Eighth?"
I think it's harder for the fans, but for the players, they'll have to be hardened for it, they've played hundreds of games in their lives, from school to college to the minors, late innings losses happen, plus they have a strong support network, professional coaches and so on, they get over it. Teams that play just once a week, football say, may get a bigger hangover
There do seem to be minor effects. But yeah, I think fans under-estimate how resilient these guys are.
come for the gory mathematics, stay for the taylor swift