Baseball Therapy: Say You’ll Remember Me

Watching a playoff game should count as aerobic exercise. You should seriously be able to go to your doctor and said “I know that you told me to go run a couple miles per week, doc, but it’s October and I’m a baseball fan.” And your doctor should simply nod. Everything is so important. Things that wouldn’t even register during the regular season get picked apart and rehashed and they get your heart racing, particularly if you’re a fan of one of the two teams playing. And there’s good reason for that. Because we’ve now entered the League Championship Series section of the competition, one little decision might be the difference between a trip to the World Series and a trip to the trivia challenge bin. (Quick, name the two losing teams in 2013's LCSs without looking…)

Managers feel it too. If you’re the type of fan who watches one team consistently over the course of a season, you probably have a pretty good feel for what moves a manager is about to make. Maybe not all of them, but it’s not like these moves are impossible to figure out. Take a simple maneuver like bringing in a reliever in the seventh inning. Your favorite manager might have a reliever whom he prefers to bring into the seventh, or perhaps he has his favorite lefty and righty options.

Of course, sometimes a reliever has a bad night. When your job description involves facing a grand total of four guys, all of them capable major-league hitters, it’s not surprising when you give up a hit once in a while. According to legend, even Mariano Rivera had a bad night once in a while. Still, relievers are often brought into situations where a hit or two can turn the entire game around. In the seventh inning if you do your job right, you get a “hold” and maybe a line in the game story about how you pitched a scoreless seventh. If you do it wrong, the game story is about how you are a terrible human being.

Worse still, the next day, your manager might think twice about calling down to the bullpen and asking for you. And in the playoffs, that nervousness factor only seems to get amplified. Managers already seem to be ordering from a smaller menu when ordering relievers during the playoffs, with the team’s best relievers often being used in every game—even when the team is up by 5—while others seem to be welded to their bullpen seat. But if one of the “good ones” has a bad day, managers might see that small menu get even smaller.

Take the case of Astros reliever Will Harris. Harris had been a mainstay of the Astros ‘pen in 2015, and pitched in the AL Wild Card game against the Yankees along with Tony Sipp and Luke Gregerson. If the Astros had their version of Hererra-Davis-Holland, it was those three. But in Game One of the ALDS, Harris had a shaky outing, facing four batters and giving up two hits. In Game Two, he faced four batters again, gave up two hits and a run, and took the loss. By Game Three, when Dallas Keuchel left after seven and the Astros were up by two, A.J. Hinch went to Sipp for two outs and then Gregerson for four. (Maybe Hinch was on to something. In Game Four, Harris was back out there, and he gave up four runs in 2/3 IP. By Game Five, Hinch was using Pat Neshek and Dallas Keuchel in relief.)

It’s easy to see why managers might get skittish about a reliever who just had a bad outing. Everything counts and maybe the bad outing is a sign of something bad to come. Maybe he’s hurt or the pressure is too much for him or he offended a billy goat in 1945. And because information that is recent (i.e., he had the bad outing yesterday) is the easiest to recall, it’s not surprising that a manager might over-weight how much importance one bad outing should be given.

Should he? Is a bad outing bad news for the next time the pitcher goes out there? If you live like that do you live with ghosts?

Warning! Gory Mathematical Details Ahead!
For the 2003-2014 regular seasons, I looked at all relief appearances and coded them as either good or bad. I used the definition that an outing was “bad” if either the pitcher had an OBP against of .500 or greater against him during his outing or if two or more runs scored on his watch (whether they were inherited or original was not important—if he inherited the runners, he failed in his job to strand them.) I excluded outings of three innings or more, but did not pay attention to any contextual factors (the score, for example). This is a loose definition, but it will do for the moment.

For each outing that a reliever made, I calculated how many days it had been since his last time out as well as whether the time before was a “bad” outing or a “good” one. I coded each batter he faced as ending in a strikeout (1) or not (0)—and the replaced strikeout with a bunch of other outcomes—and figured out the pitcher’s overall strikeout percentage, the batter’s strikeout percentage, and league strikeout percentage, and using the odds ratio method, figured out the likelihood that a given plate appearance would end in a strikeout. I used this as a control variable in a binary logistic regression and used the “last appearance was good/bad” as another predictor. Once we’ve controlled for the batter and pitcher talent level, if the good/bad indicator has any predictive power left, we know that we have something.

I selected only plate appearances where the pitcher came back the next day. In some of those appearances, he was coming off of a “good” outing, and in some a “bad” outing, but we can at least hold the amount of rest he got constant. The results: There were no outcomes (strikeouts, walks, singles, 2B/3B, home runs, outs in play, on-base events, and a couple others) where the previous outing made a difference. It’s not that a pitcher who had a bad outing before was guaranteed to have a good one this time; it’s that he didn’t pitch any differently after a bad outing (or a good outing) than his seasonal stats would have suggested he would. I looked at pitchers who had one day of rest in between outings, and once again, found little in the way of evidence that their previous outing mattered much. Same for two days of rest.

While that’s the story for the regular season, the playoffs are a different animal. So, I looked at the same analyses for the playoffs, using everyone’s regular season stats as a baseline for our expectations. We do have some evidence that regular season outcomes aren’t great predictors of playoff outcomes, specifically around the three “true” outcomes (strikeouts, walks, home runs), but that aside, there was still no evidence that a pitcher who had a bad outing last time—while it may have left his manager breathless—it didn’t leave him with a nasty scar.

He Shakes It Off
There’s a giant methodological issue that we should talk about first. I looked at what happened when a reliever went back out there following a bad outing. But the fact that he went back out there means that someone—probably his manager or pitching coach—took a look at him after the game and said “Yeah, he’s good to go.” It’s entirely possible that managers have an innate ability to look at a pitcher and realize when he’s not okay to pitch the next night or two days later. Or maybe they even have some clue, even if they aren’t perfect about it. And that’s going to bias our sample. Then again, maybe they have no idea. We can’t directly test the question because by definition, the guys that they hold back never get into the game.

We probably do over-estimate how much an unfortunate turn of events affects a pitcher. Despite the methodological issue above, perhaps we need to admit that a manager’s reluctance to bring in a good reliever in whom he has had faith all year into the same situation he’s been brought into all year is more of a reflection on the manager’s fear than it is sound strategy. The evidence suggests that if you trusted him during the regular season, you might as well trust him now.