April 7, 2014
Beware of the Intentional Walk?
I missed baseball. It’s like being in a relationship with someone and then having to spend an extended period of time apart from them. Oh sure, you call and Skype and send each other e-mails, but when you are finally back in the same room, you get the joy of re-discovering each other. (And yeah, that’s a Journey reference.) Then there’s the next day after you’ve… ahem… gotten re-acquainted, when you realize that in addition to all of the wonderful things you missed about each other, all of the things that drive you crazy are still, there too.
In the off-season, with no games to be played, there are no announcers filling the air with amusing anecdotes and half-baked theories about the game. Which means that Ben Lindbergh doesn’t send me e-mails saying, “I just heard an announcer say that [insert assertion about baseball with no evidence given]. Is that true? Can you do #GoryMath on that? Pretty please with rainbow sprinkles?” Ben knows I’m a sucker for rainbow sprinkles.
Ben’s latest rainbow sprinkle comes to us via the New York Yankees broadcast team of Michael Kay and Ken Singleton. In the eighth inning of Saturday’s Yankees-Blue Jays game, Brett Cecil was in a tough spot. With two outs and a 1-0 Jays lead, the Yankees had Jacoby Ellsbury on third and Derek Jeter (did you know he’s retiring at the end of the year?) on second. Blue Jays manager John Gibbons chose to walk out to the mound and call for Sergio Santos to face on-deck hitter Alfonso Soriano. Kay suggested to Singleton that perhaps Gibbons could have instead ordered the left-armed throwing Cecil to intentionally walk the right-handed hitting Soriano and stay in to face the left-handed hitting Kelly Johnson. Singleton replied that that wouldn’t have been a good idea, because issuing an intentional walk and then trying to pitch normally again is hard for pitchers to do. Throwing four half-hearted balls a foot and a half off the plate gets them out of their rhythm.
Warning! Gory Mathematical Details Ahead!
For all plate appearances in the data set, I coded for whether the PA immediately prior was an intentional walk. In addition, I also controlled for whether the batter and pitcher were of the same or opposite handedness (teams often use the IBB to gain a platoon advantage, as Kay was suggesting that they do by having Cecil walk Soriano to force a lefty-lefty matchup). Also, an intentional pass is commonly given to try to set up a double play when there are other runners on. The fact that those runners are out there suggests that we may be dealing with a pitcher who has thrown a lot of pitches. So, we’ll control for his pitch count as well.
As per usual, I ran a series of binary logistic regressions. The results indicate that… well, Ken Singleton was right. Sort of.
After giving out an intentional walk, pitchers strike out fewer hitters than we might expect, again controlling for the talent of the batter and pitcher, handedness advantage, and his pitch count. They also walk fewer batters and give up more singles. There was also marginal evidence (p = .13) that they give up more home runs. Ouch.
In fact, the effect size was a decrease of roughly three percentage points for both strikeouts and walks, and most of the missing walks and strikeouts became singles. Yikes. That sure sounds like a shaken-up pitcher. Victory for Singleton? Maybe not.
I’m fond of saying that 90 percent of baseball research (and probably 90 percent of the frustration that skeptics feel toward numerically-inclined folks) is figuring out how to account for bias in our data sets. The actual figuring out what a player did isn’t all that hard. The problem is that baseball provides us with a horribly biased data set. In this case, we have to think a little deeper about the intentional walk. Looking at events and asking only whether the preceding event was an IBB misses some critical information. Intentional walks aren’t just handed out in random situations. In fact, more than three quarters of intentional passes are handed out in situations where the game is close (within two runs). Also, it’s rare to see an intentional walk with the bases empty. Normally, the intentional walk is done with runners on second or third. (In fact, more than 80 percent of IBB’s from 2009-2013 were done with a runner already on second.) More than that, intentional walks are rare with no one out.
After an intentional walk, a pitcher is usually facing multiple baserunners and a close score, with an out or two already on the board. So far, we’ve compared these situations only to a more general set of circumstances, including a bunch of plate appearances when the score is lopsided or there’s no one on base. Our comparison group might not be appropriate because those additional factors might make a difference. Instead, I went a step further and started isolating situations in which there were runners on first (because after an IBB, there will always be a runner on first) and second, and one out in a close game (again, within two runs in either direction). I played around with the configuration (went to two outs, put a runner on third, etc.) and tested to see whether or not an immediately previous IBB made a difference. When I did that, a lot of the significant findings started to disappear, with some of the original findings starting to have p-values in the .15 or .20 range. Suddenly, those findings are a lot more dubious.
But of course, because I’m getting very restrictive about my inclusion criteria, a lot of my sample size disappeared, too. To correct for this, I re-ran the analyses with a database covering 1993-2013 (my computer currently hates me). It didn’t help significance very much. There was the significant finding that following an intentional walk, a pitcher gives up more singles than expected, but this probably has more to do with the fact that one reason to issue an intentional walk is specifically to set up a double play. Pitchers probably try to pitch in ways that will induce ground balls afterward, and ground balls are the seeds of singles.
Walk This Way…
The bigger lesson on this one comes from understanding that baseball presents us with a really biased data set. If we’re going to do this right, we really need to make sure that the findings from research in baseball are accounting for that bias. Intentional walks are not randomly dispensed. They’re given out in very specific situations, and once you control for those situations, it looks less and less like it’s the intentional walk that’s a problem and much more that whatever effects we see are being driven by the situation. No amount of rainbow sprinkles will change that.