Platoon splits are real, and they matter. The trouble comes in when people put too much emphasis on individual platoon performance over a short period of time. It is understandable, of course, and as fans, we have a right to overreact to things. But when it comes to getting to the truth of things, it gets a bit more complicated. One can read about the general principles of thinking about observed platoon performance versus true talent elsewhere. What can be instructive is looking at some concrete cases of unusual or standout performances of certain players. For today, let’s take a look at the platoon histories of a couple of left-handed hitters: Alex Gordon and Robinson Cano.
Gordon is having a down season with the bat. It is not that he has been out-and-out bad — after a miserable slump in June he has clawed his way back to respectability. After a big 2011 during which about everything went right for Gordon in terms of walk rate, power, and BABIP, somewhat surprisingly, in 2012, his BABIP stayed high and his power regressed. This year, his walk rate and BABIP have both dropped. But a general analysis is not what we are after here. What is really surprising is Gordon’s big reverse platoon split.
During Gordon’s first two seasons in the majors in 2007 and 2008, he showed a bigger than average wOBA split. In 2007, it was slightly larger than average for a left-handed hitter at about nine percent, then in 2008 he had a massive split of about 24 percent. Gordon dealt with injuries and demotions in 2009 and 2010, getting less than 500 plate appearances total over both seasons combined, and less than 130 plate appearances versus lefties during those seasons, a sample size that makes them basically uninteresting on their own.
During his 2011 comeback, or even “breakout” (a word I don’t like to use) campaign, he displayed a slightly smaller than average split. After that season, he had about 681 career PA versus southpaws and about a an 11 percent observed split. It was reasonable to think he had a larger-than-normal true talent platoon split, but keep in mind that for lefties, even at 1,000 career PA versus lefty pitcher, left-handed hitting hitters splits still are regressed half way to average. In 2011, lefties had an average split of just over 10 percent, so Gordon was probably around league average for that year in terms of his true talent.
Then came 2012: Gordon crushed right-handed pitching with a .391 wOBA, the same wOBA he had against righties in 2011. However, the bottom came out against lefties, and he managed just a .296 wOBA against them. His split was almost 27 percent, the biggest of his career. Even in a season in which lefties averaged nearly a 13 percent split, it was big. Did that mean Gordon was now a platoon player? He was up to 935 career plate appearances versus lefties with an observed career split over 14 percent.
Fast forward to this season. Gordon’s observed split so far is over 27 percent even bigger than last year. He has a .394 wOBA versus lefties and .303 wOBA versus lefties. It may be time for the Royals to consid– wait, what? That is not a typo. Gordon is sporting a reverse split this year. People I respect have their theories on the cause of this, and that would be important to discuss in a different sort of post. From a basic statistical standpoint, it is worth noting that after this season’s 214 PA versus lefties, Gordon’s career split now stands at just seven percent. He has 1,149 career PA versus southpaws now, so from a simple perspective (without going through the math) we might surmise that Gordon has a slightly smaller than average-sized split.
The anecdotal lesson here is perhaps trite but worth remembering given how people can sometimes overreact to one season of splits. Gordon looked like a player with a big split for a lefty, and now has a single season with such a big reverse split that, statistically, it looks like he’s about average in that respect. Yes, this season’s 214 PA versus lefties is not a big sample, but it was enough to alter the overall outlook of Gordon’s 1,149 PA versus lefties. Indeed, that itself means that Gordon’s split should be regressed equally to league average.
Even multiple seasons of a big split can look very different after one season. So one should not get too worked up about one season of a big split. Nor, from a statistical perspective, should one get too worked up about one season of a big reverse split.
If Gordon’s seasonal platoon story looks odd, Cano’s stands out even more. It is a similar story, in the end, so it will not take as long to get the “moral,” but it is worth recounting because it is a rather unusual path to the same point.
Starting closer to more recent times, in 2012 Cano had one of the best seasons by a second baseman in years. He hit .313/.379/.550 (149 wRC+, .394 wOBA), setting the stage for his intriguing offseason after this season. One concern expressed by some Yankees fans in the aftermath of Cano’s awesome season was his massive 2012 platoon split. While he had destroyed right-handed pitching to the tune of a .461 wOBA, versus left-handed pitching he managed just a .290 wOBA. This raises a number of difficult questions about how platoon splits help or hurt overall value, as well as about how they relate to player aging. Again, those are interesting and important issues, but for now I want to leave them aside and step back. Yes, it was a massive observed split. But that was not always the way Cano had hit.
In Cano’s first two seasons in the majors, he did demonstrate big splits: about 18 and 24 percent, respectively. However, in 2007, he actually managed a small reverse split over five percent. In 2008 Cano had a poor season, and weirdly had a huge reverse split of about 18 percent. One cannot simply throw that season out, and after it his observed career split was only slightly larger than average, even if it would be regressed much closer to average.
In itself, that would be a good reminder about the small-sample based randomness of platoon performance. But the tale continues. In 2009, Cano bounced back, but while he did not retain a reverse split, he did have, well, no split, at least not according to observed wOBA. He managed a .374 versus both righties and lefties, and had a close to even observed career split with over 900 career PA versus lefties. In 2010, regression still took hold, though, and Cano’s split was almost exactly league average for a lefty. Then in 2011 he again had virtually no observed split.
So, after 2011, just before the 2012 season in which he put up an absolutely massive platoon split, Cano had more than 1,400 career PA versus left-handed pitchers, and had an observed platoon split of about three percent. He had enough PA versus lefties that we would expect him to be closer to his observed split than to the average. The lesson is not to throw out our methods for estimating platoon skill. Those just produce estimates. We know that random distribution of events and the things that basically statistical methods do not taken into account matter.
But if one is not going to be ready to say that Cano had almost no split after his first seven seasons in the majors, one should not overreact to a truly massive split in one season. In 2013, Cano again has a big split, but it is nothing like last year’s. At the moment, it is about 18 percent — big, but Cano is still respectable (.340 wOBA) versus southpaws — and though his 2013 is not on the same level as his 2012, he is still having an excellent season to set up the big free agency story of the coming offseason.
How much this will or should figure into Cano’s next contract is another issue, but it does not loom as large as some might have thought a year ago. Even after these last two seasons of big splits, Cano’s careeer split is just about 10 percent. Even now, that is probably closer to his true talent split than his 2012 or 2013 performance.
Both Gordon and Cano have had seasons during which one might have surmised they were virtually platoon players given their performances, and other seasons during which they seemed to have almost no split or even a reverse split. Perhaps lessons about small samples (as one season divided up by platoons will always be) and variation within populations are old ones, but they are worth remembering. For all of their unusual moments, Gordon and Cano both end up confirming the long-term wisdom of not focusing on just one or two seasons of splits.
Print This Post