Platooning: the value for a player (part 3)

I remember how, many years ago, my father came home visibly disappointed. His cardiologist had just informed him of the results of the detailed analysis they performed on him, and the news was not good. So, my Dad pulled up a chair, sat next to me, and told me, with certain sadness that was impossible to ignore, “You know, I really feel sorry for the guy. I thought he were a good doctor.”

I remember this story every year, in the months and weeks leading to spring, as we listen to one baseball expert after another, trying to figure out how our favorite player or team will do. I like to imagine how my father would pull up that chair, sit next to me and say, “So, what you got for Chipper?”

In my dreams, my father not only is alive and actually knows something about baseball, but he also has a favorite player, some aging star who played for the same team his whole career. He would then patiently listen to me explaining park factors, splits, about dropping line-drive percentages and about aging curves, about how Chipper is just not the same player anymore, not now that he is 34. Not now that he is 36. And definitely not now that he is 40, Dad, 40 years old!

He would let me finish, and then he would say, “Chipper’s gonna hit, Son, just you wait and see.” And he would probably walk away feeling a little bit sorry for me and my faulty projection system. In his mind, he would just know that Chipper would not stop hitting next year, just like he knew back then that his heart would not stop beating within months.

If you are going against a projection, you base your expectations either on ignoring the information you have no use for or by having additional information. I am happy to inform you that my Dad beat the system using the former. I will try to show you how to do it using the latter.

One of the best ways to have an educated guess about how a player will do next year is to ask Brian about it. Brian is our own Brian Cartwright, and his Oliver projection system has been on the forefront of projecting baseball performance for years. So, I asked about Chris Young, and this is what Oliver had to say about the newest A’s outfielder. Three-twenty-one. As in wOBA of .321, a number below his previous years’ results of .325, .329 and .350.

Brian was the more talkative of the two, and he explained that a player of Young’s profile is generally a bad fit for a park like Coliseum. Young hits many fly balls, but only a below-average percentage of his fly balls leave the yard. Young also hits more pop-ups than an average player, and Oakland is the very last place you want to hit a pop-up, as it consistently ranks as a leading park in the percentage of foul pop-ups caught for outs. Adding this information to the plethora of other, Brian explains, led to Oliver predicting a decline for Young.

Now, I have no reason whatsoever to think that I can process the same information better than Oliver does, and frankly, neither do you. But imagine we had an additional piece of information that Oliver didn’t consider.* Imagine someone told us that the A’s will use Young as a platoon player.

* “Didn’t consider yet” is probably more accurate, as I’m sure many Oliver fans will be glad to hear that Brian is working on adding the splits projections to his already excellent set of information.

Now, whether the A’s will really platoon Young is not certain. However, do consider these similarities between pre-2013 Young and pre-2012 Jonny Gomes. They had virtually same wOBA splits: Gomes was at .378/.318, Young at .373/.313. Neither was heavily platooned in the past, Gomes having 33 percent of his plate appearances versus left-handers, somewhat above league average. Young has 27 percent, pretty much in line with the rest of the league. And then came 2012, and Gomes faced left-handers 59 percent of the time, turning into one of the most platooned players in the majors.

They are different players on the defensive side, so it is possible that Young, as a superior outfielder, will see more playing time. But for now, imagine that Young is not only taking Gomes’ place on the roster, but his role, too. His rate stats should be better because of that, but the question is, how much?

First, we have to estimate Young’s true talent level of handedness split. As explained in the first part of this mini-series, we do this as follows:
1. Calculate the harmonic mean of Young’s plate appearances versus left and versus right: N = 2/(1/963+1/2625) = 1409.
2. Calculate his observed split as a percentage: S% = (.373-.313)/.373 = 16.1 percent.
3. Regress with 1670 plate appearances of league-average split for right-handers: TTL = (1409*16.1% + 1670*6.1%)/(1409+1670) = 10.7 percent.

Next, we reverse engineer Oliver’s projection:
1. We assume the projection is made of 27 percent of plate appearances versus left-handers and 73 percent versus right-handers, in line with his career norm.
2. We value the true level of Young’s handedness split at 10.7 percent, meaning that his wOBA vs. right-handers should be 89.3 percent of his wOBA vs. left-handers.
3. We plug in the numbers, 27%*wOBAvsL + 73%*89.3%*wOBAvsL = .321 (Oliver projection). Therefore, wOBAvsL = .348, wOBAvsR = .311.

Now, if we mix these two numbers in the way Young’s were mixed so far (27 percent of plate appearances versus LHP), we obviously get back to .321, just as Oliver predicted. But what would happen to Young if he were to be used like Gomes was? We would then have 59%*.348+41%*.311, good for a wOBA of .333. Young’s wOBA jumped by more than ten points without any changes in the assumption of how he will bat, only about how much.

Obviously, speculations about usage are just that, speculations. But if you understand the principles of regression and use this simple approximation of projecting the split production, you can at least put a proper number on such speculations, or any other you might have when it comes to platooning.

I know, my Dad probably would say that Young won’t hit squat, except for that walk-off home run to clinch the division, and I wouldn’t really argue. But if you are like me and try to put a number on almost everything, now you at least know how.

A Hardball Times Update
Goodbye for now.

Bojan Koprivica played soccer in the USA and baseball everywhere else. While some use this fact to question his competitiveness, he talks of it using expressions like "international flair".
Newest Most Voted
Inline Feedbacks
View all comments
10 years ago

Thank you very much for this insightful series, Bojan. Any chance you could do a quick walk-through of how you would handle switch hitters? I saw you suggested 1620 PA (HM) and 6.8% split in the comments from Part 1, but I’m not grasping how they should be applied.

Coco Crisp (.322/.322) and Wilson Betemit (.281/.354) are two extremes causing my confusion.

P.S. That line from your father about his doctor is an admirable attitude.

Bojan Koprivica
10 years ago

Thanks for reading this and I’m glad you liked it.

True talent level of handedness split is much harder to estimate when we talk about switch-hitters. Sometimes it seems impossible, at least for me.

The biggest issue is that while we can estimate the magnitude of the league average split just like we do with left-handers and right-handers, we lack the uniformity of direction. We know that right-handers hit better against left-handed pitching, but we don’t know who the switch-hitters hit better against.

For someone like Wilson Betemit that should be easier to handle. We take his 834 PA(HM) of observed split of 21%, regress it with 1620 PA of 6.8% split, and we can say that we estimate his TTL to be 11.5% split, with his batting from the left side being the better of the two.

With Coco, however, things get more difficult. In his case we regress his observed level of split, which is 0% over 2073 PA (HM) just the same and we come up with TTL estimate of 3.2% But the question is – in which direction should he be regressed? Should he be better against left-handers or against right-handers?

I just don’t have an answer for that right now. I have looked into handedness of throwing for some clues, but was unable to come up with anything I could use. So, an honest answer right now is that I just don’t know how to best regress the switch-hitters who have shown little to no split, and might even be incorrect about how to regress clear-cut guys like Betemit, because determining league average out of whole population of switch-hitters might also not be the best way to go.

Finally another thing to consider is that switch-hitters not only have different pitch recognition and breaks on curve-balls when they face different handed pitchers. They also go through a completely different hitting motion. That means that some injuries or post-injury effects could affect their hitting ability disproportionately, thus skewing the picture.