October 22, 2012
Are Three-True-Outcomes Players Better in the Playoffs?
Let me pull back the curtain on how BP articles are made, at least at my house. This article came about when I was washing the dishes. It's my thing. I like to listen to podcasts and scrub down pots and pans. It's wondrously therapeutic after a long day at work, and BP alumni Joe Sheehan and Rany Jazayerli were keeping me company as I struggled mightily with the remnants of mac and cheese from my daughter's lunch plate.
Anyway, Messrs. Sheehan and Jazayerli (erm, Dr. Jazayerli) were chatting about the then-happening American League Championship Series (they seem to have recorded between games Two and Three) and pondering the struggles of Yankee hitters, even the ones not named Rodriguez. They brought up the struggles of Nick Swisher and Curtis Granderson, and Mr. Sheehan mused aloud about whether Three True Outcomes hitters (that is, guys who are particularly gifted in the arts of striking out, walking, and hitting home runs) as a class of players had trouble in the playoffs. They didn't linger too long on the subject, but at the end, Mr. Sheehan suggested that it might make a good study for someone to do. (To hear the original, listen to Episode 56, around the 1:15:00 mark).
And that someone just happened to be cleaning cooked rice out of a pot when he heard that irresistible dog whistle. My life is as glamorous as it gets.
As always, if you're afraid of numbers, please skip to "What It All Means".
Warning! Gory Mathematical Details Ahead!
In the postseason generally, strikeout rates spike, generally at the cost of singles, double/triples, and outs in play. There's an interesting question to ask here. Are the differences because the quality of the players involved in the postseason is better (and the good ones play more), or is it because players change their approach in the postseason?
I've added two columns to the above chart. One shows what we would expect the distribution of outcomes to be based on the batter's regular season results (weighted for actual playing time). The second is the same for pitchers.
Postseason rates seem to hew a little closer to the pitcher's side of the ledger. But let's take a look at batted balls.
Here we see that in the postseason, the results that we get are nearly identical to what we would expect from the pitcher's stats. This is evidence for (although I should stress not proof of) the thought that in the playoffs, good pitching really does have an upper hand. But let's see if these results change based on the batter's tendencies toward Three True Outcomes.
I started with all plate appearances from the 1993 through 2011 playoffs. For each plate appearance, I assigned the batter and pitcher their regular season stats. I also only looked at plate appearances where the batter and pitcher had both logged at least 250 PA or batters faced during the regular season. (You get the occasional Pete Kozma, who was a late-season call up and logged only 82 PA during the regular season.) This still left more than 34,000 plate appearances.
I calculated seven rate stats for each batter and hitter, including the rate at which a plate appearance in which he was involved ended in a strikeout, walk, hit batsman, single, double/triple, home run, or out in play. I excluded all cases of intentional walks (sorry, Barry) and pitchers batting. From 2003-2011, I also calculated the rates at which batters hit fly balls, grounders, or line drives.
Using a method I've described before and the regular season stats, I generated the expected chance that each of the seven outcomes listed above would happen. This ensures that we’re accounting for the fact that better hitters (and better pitchers) make the postseason. (For the initiated, I took the natural log of the odds ratio of the expected probability.) I also calculated the batter's TTO percentage, again from the regular season.
I ran a series of seven binary logit regressions. For each plate appearance, I modeled whether it ended in a strikeout. For that regression, I entered the natural log of the odds ratio of the expected strikeout rate, and then the batter's TTO rate. (Again, for the initiated, I entered it raw and then did a natural log of the odds ratio. Same results, either way.) I then repeated for each of the other six outcomes, checking to see whether TTO percentage was a significant predictor of any of those outcomes.
But by how much? I need to construct something of an artificial set of circumstances. There's a spherical cow in here somewhere, but direction before precision. We're going to assume that our hitter is 2011 league average, and that he is facing a 2011 league average pitcher. We're going to vary only how much his TTO percentage changes. (Somewhere out there, someone is thinking "Wait a minute, if he's league average in all respects, then he will have a league-average TTO rate. Plus, aren't you completely ignoring a whole bunch of covariances?" Yeah, I know.)
I calculate expectations based on TTO percentages from 20 percent to 40 percent (in 2011, Jeff Keppinger had the lowest TTO percentage among qualifiers at 11 percent percent; needless to say, the leader was Adam Dunn at 53 percent). The idea is to get a basic idea of the magnitude of how much TTO changes outcomes in the playoffs.
Predicted outcomes for this matchup:
The numbers are a little off, partly due to rounding and partly due to the methodological flaws in the model that generated them. I wouldn't recommend summing those across rows. They do however, show the magnitude of the effect of TTO percentage rather well. Going from 20 to 40 percent TTO, we're talking about a percentage point or two on most outcomes, again over what is expected based on the batter/pitcher matchup.
Let's look at batted balls in the same way.
Here, we're seeing a pronounced swing to fly balls over grounders, on the order of about two percentage points. Players who tend toward TTO during the season are generally fly ball hitters. Even though they are facing a suite of pitchers who are much more groundball happy, our TTO hitters put more balls into the air than would be expected of them.
What It All Means
The question of why this happens is tricky and enters the neighborhood of rank speculation, but here goes. The increase in fly balls suggests a little bit more of a tendency to aim for a home run (and apparently, with some amount of success). The increase in walks could be the pitcher recognizing this and being more careful. Alternately (or perhaps concurrently), TTO hitters also tend to be rather patient chaps in general, and they may just be even more selective in what they swing at during the postseason—but when they swing, they try to hit it a little further. It's not an all-or-nothing approach, just a little shading of preferences.
Why is it that TTO hitters don't seem to come through during the playoffs? It seems that the answer is that they are following a slightly higher risk/higher reward strategy in their plate appearances (higher than the usual TTO approach). Over a small sample size like five games, a high risk/high reward strategy is going to flame out a lot of the time.
There's the oft-quoted line from Billy Beane about his excremental ineffectiveness in the postseason. There has always been a question of whether certain types of players are better built for the long haul of the regular season than the sprint of the playoffs. The A's always seemed to have trouble in the playoffs. Perhaps the players that Mr. Beane preferred were defective when it came to a best-of-seven series?
Are there certain types of players that perform better in the postseason than we would expect? In this case, we see that TTO hitters seem to become more extreme versions of themselves. This produces detectable differences in outcomes, even if they are small in magnitude compared to the general differences between players that we normally see. And they needn’t necessarily make hitters better or worse—they may simply change the risk profile that goes along with determining the winner of a best-of-seven series.
But there's always a hidden lesson. I think that there's an assumption among the sabermetric Twitterati that the playoffs can be modeled as simply an extension of the regular season. These results suggest otherwise. The postseason really is a different animal.