More photos » Ben Margot - AP

Matt Holliday hitting a single off of LHP Jon Lester -- is this regression to the mean before our eyes?

Browse more photos »

Everyone knows that that left-handed hitters generally hit better against right-handed pitchers, and right-handed hittesr hit better against left-handed pitchers. Still, players actual career/last season splits are so varied that we sometimes latch onto those -- "hey, this righty actually hits RHP better than LHP! He's no platoon candidate," or, "this lefty-hitting rookie is so bad against LHP, he's a career platoon guy." As you might expect, much of this is generated from small sample sizes. But we also have the correct intuition that, although the general rules about platoon splits hold true, each player has their own particular split in terms of the "direction" and size of the split. With that in mind, I have taken a group of players and applying a simple method roughly following a suggestion from The Book in order to project what their "true" splits can be expected to be going forward. At the end of the season, we'll check in and see how the method did.

I initially planned on doing this in a very "precise" manner with weighted averages for each player, dividing by leagues, etc., but my first forays were such a pain, and the potential marginal "reward" so small, that I decided to go with something more simple, and in a sense, that's even more interesting, since we are dealing with regression to the mean. Here's the "clue" from The Book (page 163) that I'll be following:

A right-handed hitter needs around 2,000 plate appearances against left-handed pitchers before his measured platoon spltis can be reliable (in other words, using the measured platoon split is more accurate than assuming has an average split). From a practical standpoint, right-handers are best assumed to have average platoon skills, unless one is willing to make the calculations needed to accurately estimate a player's platoon skill. For lefties, the number is about 1,000, which means that only veteran starters have reliable platoon splits.

What I've done is simple enough. Taking a couple ideas of my own and some reader suggestions, I got a group of players with varying sorts of splits, and simply took their career splits from baseball-reference.com. I did this on Monday, so I'm assuming that no one will think that one week of 2009 stats is going to "contaminate" this oh-so-precise study. Next, I took a weighted average of the average split from the AL from 2005-2008 from the league pages. Why just the AL? Well, I first did all of MLB, but the large number of pitcher PAs from the NL makes the numbers really weird... If I knew a way to get "double splits" with position and handedness out of B-R.co or even my BDB database, I would have done that, but for now, well, we're going to have to use the AL splits for everyone... sorry, NL folks... The AL pitchers do have some PAs, but it's not that significant.

The stats I am going to use for this study, for each player, are on-base percentage (OBP) and weighted On-Base Average (wOBA --the original version, for ease and consistency of calculation). I will be projecting the percentage of each player's split, as 1) this isn't a column about projection in general, and 2) as noted in The Book, better hitters generally have larger splits.

Here are the weighted (10-9-8-7) averages I calculated for the 2005-2008 AL, expressed as percentage of the split relative to average OBP/wOBA.:

As I said before, the subjects for this study were suggested mostly by readers, with a couple of suggestions of my own. Alex Gordon is a young left-handed hitter with a typically large split -- it will be interesting to see how that works itself out as he (homerism alert) destroys the AL Central, especially Cleveland, over the next few years. Miguel Olivo hits like a catcher who doesn't hit very well, but he has mashed lefties so far in his career. Will that continue? Joe Mauer is pretty good, I guess. He was suggested because he had a reverse split in 2008, although that isn't typical of his career. Ichiro! has a "reverse" split for his career, although that hasn't persisted as much, lately. Ryan "Out of Nowhere" Ludwick and Evan Longoria are righty-hitters with very big reverse splits, as is Matt Holliday although his is not as large. Akinori Iwamura and Ian Stewart are lefties with reverse splits -- in Stewart's case, it is really big (although in a small sample), and this was also the case in the minors (I've only used major-league numbers here). Here's the data with an explanation of the columns following.

Table Key:

• PA is career plate apperances vs. LHP only.
  
• OBPspl and wOBAspl are the career OBP/wOBA split of the player. Negative numbers indicate a reverse split.
• OBP% and wOBA% are what the career split as a percentage. Negative numbers indicate a reverse split.
• xOBP% and xWOBA% are the expected "true" split percentage. Negative numbers indicate a reverse split.
• Rel. is a sort of like a Marcels reliability score, in that it is based on the number of PAs vs. LHP the player has as a percentage of his total Pas vs. LHP + PAs of lgAvg. They should be taken as purely "relative" to one another in this context however, not as anything like a "percentage chance this projection is right." I don't know how helpful they are. The reliability scores of lefties and righties should only be compared to one another, as well, since 1) the are regressed a different amount, and 2) there is a greater variance of platoon skills amoung left-handed hitters than right-handed hitters. I thought about doubling all the right-handed Rel. scores, but I'll let people figure it out. I thought about leaving them off entirely; I hope they don't distract from the main point of the piece.
  

There you have it. Those are some rough projections for the "true" platoon splits of these players based on available major league data. Note that most of them are "curious" cases, for what it's worth. Also note that all of these players coming very close or being way off of their projections doesn't either "prove" or "disprove" the theory -- 10 players over one season is a ridiculously small sample size. I'm just interested to see how it turns out. I didn't get as specific as I'd originally wanted, but it's a start. If most of them end up closer to the projected split rather than their career split, I'll be pleased.

Now I'll have to remember to check back in on these guys at the end of the season...

Update April 16: changed some labels to make things a bit less confusing.