 |
||||||||
|
|
|
|
|
The Myth of Lefty
Mashers I apologize to many of our Baseball Primer readers, as this article is more or less a rehash of a discussion that went on there over the last few days. It's not my work, and I don't pretend it is, but I hope this article enlightens people who are unfamiliar with the work about this research, and it's also handy for me to condense this into an article of a few pages, in order to try to better understand it myself. Rob Neyer also touched on this work, but his article was one of his weaker ones, in my opinion, as difficult as it would be to condense this material into his column space he did not do a great job explaining their work. He also came out with another article on this as I was writing this piece, but at this point I'm not going to abandon this, so Rob wins the speed war. Remember the first time you heard about defense independent pitching stats (DIPS)? The theory that pitchers had no control over batted balls in play, save home runs, blew my mind, and I'll bet it did for many of you, too. This was so counter-intuitive to the way many people viewed the game that many people did not believe it. Others, like myself, viewed it with some skepticism, thinking that Voros McCracken might be right in that pitchers had less control than we thought, but they still must have had some control. Revised studies of DIPS have shown that there are a few general exceptions to DIPS, such as knuckleballers and extreme groundball pitcher, and some players demonstrate the consistent ability to give up more hits than the league average, or less in other cases, and that while Voros potentially underestimated pitchers ability to some degree, he was still correct in that, by and large, they had a miniscule effect on balls in play; far less than most people thought. Well I found out about some more research in a similar vein this weekend at Baseball Primer, apparently originally done by Bill James and expanded upon in the book Curve Ball. While it is probably not as far-reaching as DIPS, it still proves a taken-for-granted belief that many people held is false. Apparently lefty-mashing, that is the ability possessed by some right-handed hitters to effectively hit left-handed pitchers, does not exist as a skill. To quote Bill James is his 1988 study, "the platoon advantage is real and virtually universal." There is a platoon advantage that right-handed batters (RHB) gain over left-handed pitchers (LHP), compared to when they face right-handed pitching (RHP). However, the key lies in his statement "virtually universal", that is, there is no skill involved in hitting LHP, and given enough at-bats everyone will regress, or rise up (if they have a history of struggling against LHP, to the mean. Weird splits against LHP, such as for example, those shown by Ken Harvey last year (and I'm unsure why I didn't mention the small sample size of 156 at-bats in my little piece on Harvey a few weeks ago), are merely the matter of chance, and given enough at-bats Harvey will revert to the average platoon splits. Now, keep in mind these results are only true for RHBs, and left-handed batters (LHB) do not subscribe to this theory, and I will go further into this point further on. If you want to find a good RHB to face lefties, you shouldn't predict his future success against LHP based on his previous plate appearances against LHP, you should look at which one has proven to be the best overall hitter, given plate appearances against pitchers of either hand. A batter will hit lefties well if he is a good overall hitter, there is little evidence to suggest that a batter's past success against LHP is a matter of skill. Furthermore, even if it is, it is a skill possessed by such a small percentage of MLB RHB that it cannot be accurately measured, and can be assumed not to exist. The average platoon ratio is 1.09. What that means is that the average RHB will perform 9% against LHP than he does against righties. To take a quick example to demonstrate what this means, Vernon Wells last year had an OPS of .892 against RHP. Therefore, one could have expected him to have an OPS of .972 against LHP, when he only had of .957. In fact, for his career Vernon's platoon ratio is 1.04. Rather, than this demonstrating any sort of small, but innate, difficulty hitting lefthanders, one could reasonably expect Vernon's batting against LHP to improve in the future. Many of you may be asking what evidence we have of this nature that backs up this claim that everyone will revert to the mean of 1.09? How do we know that 1.09 isn't just the cumulative sum of hitters like Mike Lowell (2,471 career at-bats vs. LHP, career platoon ratio of 1.01) and Miguel Cairo (2,012 career at-bats vs. LHP, Career platoon ratio of 1.18)? Again, this is not my research, but was done by some folks over at Baseball Primer, most notably a poster named MGL. The evidence begins to show itself when looking at RHB with any number of plate appearances from 1998-2003. For every even year players with a platoon ratio greater than 1.09 were looked at, and then compared to their appearance in the next year, the odd year. 375 Batters 96,925 PA's in year X with an average platoon ratio of 1.2300 122,945 PA's in year X +1 with an average platoon ratio of 1.0867 Now, the same study with batters who had a platoon ratio of less than 1.09 458 Batters 99,219 PA's in year X with an average platoon ratio of 0.8992 142,591 PA's in year X + 1 with an average platoon ratio of 1.0870 Regardless of how well, or how poorly, the batters, as a whole, performed in the first year, they showed a remarkable ability to revert to the platoon ratio of 1.09 in the second year. Now, its quite possible that there exist some players out there with an ability to hit LHP better than RHP, at say a 15% rate, or such. With regards to any human activity, it's always going to be true in my opinion that certain people are better at it than others, even to a miniscule degree. The question becomes whether or not this skill is detectable, and then if it is detectable, is it significant? The evidence above helps to demonstrate that, for the baseball population as a whole, this does not appear to be a predictive skill at all. Another poster, admittedly expressing some doubts about the theory when he looked at the recent success against LHP of players like Matt Williams, Julio Franco and Eric Karros, took further steps to show that this, even as applied to those with extreme splits, does not seem to indicate any observable skill. He took the top and bottom 20 batters, with regards to their platoon splits, with 100 PA against LHP in 2002, and at least 85 PA against LHP in 2003. What he found was that 16 of the 20 high splits had their splits go down, while all 20 with low splits had their ratio increase. Seven of the top 20 splits went below the average ratio of 1.09, while 15 of the low 20 splits at least tied the average ratio of 1.09. None of the bottom 20 fell below their 2002 ratio, while only 3 of the top 20 improved upon their ratio (Moises Alou, Marquis Grissom and Eric Karros). Clearly, there does not seem to be the same correlation here you might find in a category like home runs, stolen bases or pitcher strikeouts. Interestingly, all three of the players to improve upon their ratio, Grissom, Karros and Alou, are all elderly ballplayers, similar to Julio Franco and Matt Williams, both of whom posted noticeably extreme splits towards the end of their careers. I do wonder if there is some ability that some elderly RHB gain which contributes towards them being able to hit off LHB at a rate exceeding the average. It is a very small sample size, looking at a few of these players, but they are all well over 30. However, even if this an ability gained somehow by RHB, and it has to be gained at some point in their career, because the career splits for most of these players are just fractionally above the norm (exactly what'd you expect when you consider an 'average' career with two or three years at the end when they mashed lefties), then can a baseball club determine which older RHB gain this ability, and which ones are posting extreme splits due to a random season or three? I think that would be very difficult, so even if there is a small truth to elderly batters learning something about hitting LHP, it still doesn't really change the crux of the theory. Some people may point to outliers such as Mike Sweeney, Aaron Boone, A-Rod, Preston Wilson, Jeff Kent, David Bell and Doug Glanville (combined 26,681 at-bats, all have ratios under 1); as well as outliers such as Manny Ramirez, Benito Santiago, Frank Thomas, Brian Jordan and Reggie Sanders (combined 28,096 at-bats, all have ratios of 1.14 or above) as proof that some players are exceptions to the rule. These players were taken, to show a sampling, from a larger list at Baseball Primer. However, to directly quote MGL on what I think is a very good point, "the so-called outliers mean nothing in and of themselves. The question is whether the number of outliers reasonably conform to the number of outliers expected by chance alone, if every player had the same ratio." To draw one's attention to the event of flipping a coin a hundred times in a row, when doing such a thing, it is "usual" for one to have a long run of between 6 and 8, of heads or tails, in a row. It's unusual, but not terribly uncommon, to have the longest run be as low as 4, or as high as 10. So in flipping a coin 100 times in a row, it's not statistically significant if 10% of your tosses in row land on one side or the other. That speaks to the role that chance plays in any random event. Given enough time, these outliers should revert back to the norm, and in analysis it's prudent to believe they will, because, even if some won't, you can't separate that miniscule percentage from the general population. Furthermore, you can't even tell it they aren't reverting because they just haven't had enough at-bats, or if they have some skill. MGL goes on to talk about standard deviation and regression to the mean, which I have a rudimentary understanding of, but not enough to attempt to prove him wrong, if I thought he was, which I don't. He eyeballed the chart of the outliers and said that if they were selected from a list of around 120 possible players, than that is statistically consistent with those expected by chance alone. So, to conclude 1) On the whole, if you want to predict how an MLB RHB will perform against LHB next year, the best way to do that is to take is stats vs. RHP and multiply them by 1.09, rather than look at his stats vs LHP. There maybe exceptions to the rule, but it seems impossible to determine which exceptions are due to chance and which are due to skill, and since the vast majority will be due to chance, even if a few possess the skill, it is best to assume that any difference from the average split is due to chance. 2) This does not work for LHB, or at least it doesn't work to the same degree. LHB have a platoon rate of 1.20, to begin with. Now, it is still better to use the league average platoon ratio if one is looking at anything less than 3 years of stats, however, once you being to deal with four or five years of stats and there is a noticeable difference from the expected platoon, than you could think that is based on ability. The reasons for this probably stem from the fact that, since RHP are predominant at any level of baseball, it is all but impossible for a RHB to reach the major leagues without developing the ability to hit RHP to a degree acceptable in the major leagues. Players who simply cannot hit RHP get left behind at college or in minor league baseball at some point. However, LHB are in a different scenario, as it is quite conceivable that, because they face limited LHP, they could simply never develop the ability to hit LHP, but can still hit RHP. A RHB who can't hit RHP will never make the majors, but its very conceivable that LHB who can't hit LHP could make the majors, and it happens all the time. In other words, while Toronto fans can expect Vernon to hit LHP better next year, Minnesota fans shouldn't expect anything but sucking from Jacque Jones against southpaws. I had always assumed that a batter had two separate skill sets: hitting RHP and hitting LHP, and the news that RHB basically only have the ability to hit pitching as a whole is very surprising to me. If this article has been unclear, or you'd like to read more mathematically in-depth discussions of this discovery, visit the thread at Baseball Primer. Thomas Ayers welcomes feedback on this or any other article he has written. Also, I'm going to make some political commentary here, so be aware. Guess how many times CNN played Howard Dean's "Scream" Iowa speech in the four days after he made it? The answer is 633. Think about that. That is more than once every ten minutes for FOUR STRAIGHT DAYS. And people wonder why he lost momentum so fast? Once every ten minutes his speech was misrepresented on CNN, and that's not counting other networks, for four straight days. He came second in Washington though, and he's still going strong. Kucinich came in third there, too. What a state. Devoted Andy Brown fans need not fear, he and I traded places this week and his column should be up Thursday. Sorry for the downtime last week, we had other matters to attend to. We all apologize.
|
||||||
|
©Copyright 2003
Phil Orr
|
