Notice: Trying to get property 'display_name' of non-object in /var/www/html/wp-content/plugins/wordpress-seo/src/generators/schema/article.php on line 52
keyboard_arrow_uptop

Nothing grinds my gears within the realm of sabermetrics as much as the avoidance or misuse of context. For whatever reason, something simple like using the appropriate frames of reference has become the type of analytical tool that makes sense to implement only when it furthers a predetermined point. In fact, if nothing else is taken away from this article, heed the following warning: never trust any conclusions drawn from performance splits if the league average of that split is absent from the same analysis. If the goal involves evaluating pitcher performance with the bases empty vs. with ducks on the pond, the results of each individual must be compared to the aggregate league split before anything of interest can be parsed. The same can be said for gleaning useful information from platoon splits, which happens to be our topic du jour. Platoon splits are certainly existent, but before suggesting that certain players should split duty at a position, we need to understand what the data really explains and the appropriate context in addition to potential shortcomings for which to look out.

At its deepest root, platooning refers to using multiple players to man a position. Nowadays, ‘multiple’ can be more accurately represented by el numero dos, as platoons primarily consist of two players providing the production that other teams seek from one individual. These platoons are generally contingent upon handedness; the right-handed hitter faces southpaw pitchers, and vice versa. From an overall standpoint, studies have proven it undeniable that hitters perform better against opposite-handed pitchers than their same-handed brethren. From 2003-08, here are the yearly figures against same-handed pitchers:


Year SHP-REqA  OHP-REqA
2003   .757     .793
2004   .770     .797
2005   .752     .786
2006   .767     .807
2007   .757     .801
2008   .751     .789

The data clearly indicates that, overall, hitters fare worse against pitchers from the same side. Split even deeper, though, and the effects are more pronounced amongst lefties. One line of reasoning that surfaces quite often with regards to the more substantial splits for lefties offers that northpaws account for approximately 70 percent of all plate appearances; while righties get to see same-handed pitchers with great frequency (allowing adjustments to be made), lefties are not necessarily privy to the same opportunities. With such relatively few plate appearances against same-handed pitchers, the lefty-on-lefty matchup is a bit more foreign than its right-handed counterpart. Platoons also become somewhat difficult to implement with regularity because of the current fashions in roster space assignments; given the choice, teams will opt to employ the talented solo artist as opposed to the dynamic duo. In Strat-o-Matic leagues (especially the superstar league in which I partake), such dynamic duos are much more relevant, because a well-executed platoon can come close to creating a makeshift Pujols.

Having covered the background of platooning, there are a few major points to hammer home, the first of which reiterates my earlier plea: players must be compared to the league when discussing performance splits. The league… not themselves… which is a point many seem to gloss over. Just because Player A posts an 1100 OPS vs. opposite-handed pitchers and a 750 OPS against his same-handed opponents does not automatically trigger an invitation to the platoon party to be mailed; this hypothetical player must first be compared to the league in both areas. Comparing a player to himself tells us nothing of importance other than his own trends. It fails to incorporate context, and we are left wondering how the 1100 stacks up in contrast to the rest of the league’s performance against opposite-handed pitchers, and how that 750 OPS against same-handed hurlers compares to the rest of the league against pitchers from the same side.

All of which brings us to Ryan Howard, a popular candidate mentioned in platoon discussions in several different venues by a number of different columnists, including on this very site. My goal in discussing Howard is not to refute anything previously written, since some of the underlying logic makes a lot of sense, but rather to introduce the appropriate context to the discussion and then see what the data indicates. Essentially, those of the opinion that the former MVP should be platooned point out his drastically different handedness results. And they would be correct, as Howard has a .303/.409/.657 line against righties, with a triple-slash performance of .227/.308/.452 against lefties. When birthday-suited up, these numbers clearly lead to the conclusion that Howard stinks against same-handed pitching.

Using Howard himself as the context, however, misses the boat and instead works similarly to price anchoring on infomercials. The commercials will display a product and explain how it used to cost $100 but can be had now for just $35; stacking up the $35 to the $100 is not the correct basis for comparison, as the $35 should be compared to the current price of other, similar products elsewhere, and not to its reported former price. After calculating the slash lines for hitters against same-handed pitching in each season, here is how Howard has fared relative to the rest of the league:


Year   PA  SHP-REqA  lgSHP-REqA  Delta
2005   63    .429      .752      -.323
2006  225    .925      .767       .158
2007  246    .839      .757       .082
2008  265    .762      .751       .011

Self-context suggests that Howard is a perfect platoon candidate. The appropriate context pegs Howard as well above-average against opposite-handed pitchers, and at worst, a league-average hitter against the same-handed. As was mentioned before we got into Howard’s specifics, some of the logic to platooning a player who hits like this makes sense. After all, first base is a premium offensive position, and the goal should be to maximize production when possible. If a team gives 70 percent of its plate appearances to Howard-since righty pitchers accrue that percentage of playing time-it stands to reason that a suitable sidekick could be found to handle the 25-30 percent of plate appearances with lefties toeing the rubber.

However, with the full knowledge that Howard has been above-average or average against same-handed pitchers, might there be better platoon choices on his own team, let alone the rest of the league, especially given the expense of using a roster spot on a platoon partner? The table below shows current Phillies starters who don’t switch-hit, and their Raw EqA (sans stolen-base components) against same-handed pitching in 2007-08, as well as the same metric for the rest of the league, and the differential:


Hitter        SHP-REqA lgSHP-REqA  Delta
Chase Utley    .911      .755       .156
Jayson Werth   .783      .755       .028
Raul Ibanez    .779      .755       .024
Pedro Feliz    .699      .755      -.056

Howard may be trending downwards against same-handed pitchers, but the table indicates that he might not even be the player best suited for shared duty on his own team. This segues into the next major point of contention-the data simply isn’t as reliable as we think. The Book: Playing the Percentages in Baseball dedicated an entire chapter to this topic, and Dan Fox followed up on the point in 2006, but allow me to summarize some of their findings. First, platoon splits are normally distributed, lending themselves more to chance than individual skill. When a dataset is normally distributed, in the form of a bell curve, we cannot say that incredibly good or bad marks are not the direct results of chance, since such distributions call for certain percentages of its sample to exceed two or three standard deviations of the mean. This is not to say that all platoon splits are derived from chance, but rather that separating skill from chance proves to be a much more difficult task, mainly due to sample-size issues.

With such small plate-appearance samples accrued each season, investing in the stock of platoon splits is not very prudent. Dan used split-half reliabilities relative to career handedness splits to show very unstable performance, and my favorite toy, the intra-class correlation, concurs. If we usually require 1,500-2,000 plate appearances over three years from a player before determining a semblance of his true talent level, why on Earth would 125 trips to the dish in three straight seasons against a specific side of the mound convince us of productivity levels? Simply put, platoon splits are consistent in the aggregate, but not from an individual perspective. Because of this fact, in conjunction with the sample-size issues, Tango, Lichtman, and Dolphin concluded that the platoon split of an individual player had to be massively regressed to the mean to get an accurate indicator of the true differential. For righties, 2,200 plate appearances worth of regression needed to be factored in, with 1,000 plate appearances for lefties.

Overall, platoon splits might be the byproduct of chance, are accrued in small pools of plate appearances incapable of determining statistically significant consistency, and must be heavily regressed before any real talent level can be deciphered. With these characteristics in mind, I fully support the suggestion of relying more heavily on scouting information than the numbers to determine a hitter’s ability against same-handed pitching. If a scout can note a specific player struggling in terms of bat speed or perceived pitch recognition against a same-handed pitcher, and the evaluations are corroborated by a plentiful bounty of trustworthy sources, I would be more inclined to base decisions off of this information than a heavily-regressed, potentially inconsistent, platoon split. The takes of legitimate scouts can indicate the roots of certain struggles or changes in approach against same-handed pitching, the sorts of thing that could lead to upticks or downturns in a player’s performance. The best information is the most accurate, and in this case, scouting information likely has an advantage over the data.

Platoons certainly exist at the major league level and, in the aggregate, hitters definitely perform better against opposite-handed hurlers, but the data used to evaluate such performance needs to be viewed with the appropriate contextual lens-the league, not the individual. Even then, it might not explain what we think it does. The bottom line? Be careful when it comes to platoon splits and assign heavy doses of skepticism about small samples of solid or awful performance.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
mnsportsguy1
7/03
Eric, your work is one of the main reasons I subscribe to BP. Another great article. Keep up the good work.
EJSeidman
7/03
Appreciate it!
jsherman
7/03
While I like the context-sensitive approach in many areas, I don't think it works in platooning. First, I think you need to keep separate the average stats for a lefty-lefty and righty-right matchup as the appropriate context, since the deltas are so different.

Next, don't we need to adjust for position? A shortstop with a .700 EqA vs lefties is far different from a first baseman with that number.

Second, your chart showing same hand delta EqAs for the four Phillies implies that Feliz is the best candidate for a platoon. But really all the chart shows is that Feliz is is below average hitter, and Utley is a really good hitter.

Why would you compare Howard vs lefties to other lefties vs lefties to see if he should be platooned? If every lefty posted .600 EqA against lefty pitchers they'd all be average, but you'd still want to platoon many of them, especially at easy positions like first base.

Even if Howard is above average compared to other lefty-lefty matchups, all you need is someone who can play first base and hit significantly better (due to the cost of using an extra roster spot) than Howard vs lefties, then you should platoon them.

Of course, the points with regard to sample size and platoon split stability are spot on and make platoon decisions quite difficult.
ostrowj1
7/03
Shouldn't the answer always revolve around who you are platooning with? Even if Ryan Howard were exceptional facing lefties, if you had (or could acquire easily) a 1B who was (significantly) better against lefties wouldn't you want to do that (enough so to justify the roster space)? I don't see what positional adjustment or comparing splits to the average hitter have to do with anything. My guess is Howard may be a better platoon option than Feliz mostly because it is probably easier to find a lefty-masher who has a reasonable enough glove for first than a third baseman. Obviously, Feliz is a better candidate for replacement...
EJSeidman
7/03
Essentially, to take into account above propositions, you don't need to use data to decide who to platoon. If legitimate scouts are indicating a player struggles in certain circumstances, that is likely more accurate than small samples of data.

HOWEVER, the point I'm making here is that many of those opining that Howard should be platooned point to his drastic self-split, which is not the appropriate context. If the Phillies are trying to decide which player to platoon, they should be factoring in defense, what diminished playing time might do to the player against OHP, and which of their players is both easily platoonable from a personnel standpoint as well as a viable candidate given their ability to hit SHP compared to the rest of the league.

So basically, you don't need to use data or comparisons to decide who to platoon... but if you're going to use data, use the appropriate context, which is some form of the league against same-handed pitching, not individual splits.
ostrowj1
7/04
I understand what you are saying. I read the article as "things that need to be considered when choosing a platoon" instead of "don't bash Ryan Howard because of his splits" (maybe that is simplifying things just a bit).

A further not on the small sample size issue... There is also the concern that a players like Howard (whose "stuggles" against OHP are well known) will often get days off when ace (or merely good) OHPers take the mound. Thus, even though his numbers maybe decent, they may be against weaker than average competition, adding more a lot more variance...
EJSeidman
7/04
That would honestly be evened out by teams throwing their top specialists at him much more often now. Or making sure to use a lefty reliever when he comes up when they might otherwise not. Plus, there aren't many elite LHP starters. And Howard usually plays vs. Johan.
EJSeidman
7/03
Quite honestly, the LH-LH and RH-RH were eerily similar. Thus why I combined them, there was really no difference. LH-LH would have a .754 and then RH-RH would be at .757 or something like that. You COULD adjust for position and such but the idea of showing this data was merely to indicate that the appropriate context ISN'T the player himself but the league.

All sorts of scouting and makeup issues come into play, though. For instance, would Howard NOT playing everyday, as he is used to, adversely affect his awesomeness vs. righties? Players are not strat cards and while the numbers are more accurate in many areas, we're both in agreement that sample sizes prevent accuracy in this area and that scouting is probably much more important.
blcartwright
7/03
Here's another way to look at it.

Marcel estimates a players 'True Talent Level' by taking a weighted mean (5/4/3) of the last three seasons. Then apply the same weighting to the difference of how the hitter does vs rhp and vs lhp, producing the split delta.

Say a hitter has an overall .360 wOBA, a split delta of .060, and 2/3 of PA vs rhp, 1/3 vs lhp.

(2/3)*R + (1/3)*L = .360
R - L =.060

Solving that gives us that the player has a .380 wOBA vs rhp, .320 vs lhp.

Repeat for every hitter on the roster, and then find your best lineup vs right, and best vs left.
ostrowj1
7/04
Why is doing this better than just taking the weighted mean (adjusted for # of at bats) of the wOBA vs left or right handed pitchers? (I don't have a pad of paper handy, maybe it is exactly the same). Either way, this way seems to be needlessly obscuring things, but maybe I am wrong.
blcartwright
7/04
I did neglect to specifically mention that Marcel and projections in general use regression to the mean, based on the number of PAs in the sample.

If you run vsR and vsL seperately, you are going to regress each of them, so twice for that player. Even though you would regress the overall batting line and also the split delta, the split only affects how much is allocated to vsR or vsL, not the overall total. The idea for this process comes from "The Book" that Eric references in the article, and we've had some discussions of it a few months ago at Inside The Book Blog.

In this method, you only need three pieces of data - the weighted mean of all batting, the weighted mean of the split delta, and the allocation of PAs between vsR and vsL to get the estimate against each hand.
jnossal
7/04
Great article, Eric.

Speaking of context, I'd like to point out that in all likelihood many of the sabremetric writers advocating for platoons, Howard in particular, participate in Strat leagues or something similar. You even alluded to this in your article. The problem with that is those leagues are based on known stats for a past season, making platooning choices a simple matter of mathematics. But teams playing this season must make decisions in real-time, without any prior knowledge of the likely outcome other than the noted sample-size limited near past.

It is this context I believe that often drives these arguments. The writers may have their premise somewhat skewed based on Strat league roster construction while neglecting the real problems faced by actual teams. I'm sure it is an unconscious bias, but likely a very real one.
Oleoay
7/04
I haven't heard of teams doing it since the mid 80s, but I vaguely remember a platoon not based on handedness, but on how well batters handled fastball or offspeed pitches.
JoshC77
7/06
That's an interesting point. When tying this into the discussion of handedness, could it be tied into the fact that lefties tend to be softer throwers with more off-speed stuff and there are more righties with higher velocity stuff? I realize that this is a gross generalization, but it's just food for thought. It would be an interesting study to see 'platoon splits' of current players versus hard and soft throwers.

On a side note, this reminds me of former Reds 3B Willie Greene. He could mash a fastball but was clueless against breaking balls (think Serrano in Major League). I remember one series versus the Cubs where they had shut him down the first couple of games then inexplicably started throwing him fastballs. End result....lots of bombs.
blcartwright
7/04
Platooning was a lot more common in the 1970's and 80's when most teams had a ten man pitching staff, leaving seven on the bench. The '79 Pirates used Nicosia vsL and Ott vsR at catcher, and Robinson vsL and Milner vsR in left field. Iy was a strict platoon on hand of the pitcher, everyone knew ahead of time which games they would start.
EJSeidman
7/04
Right, and these days, with 12-man pitching staffs and one of the four bench spots reserved for the backup catcher, and another seemingly for a fourth outfielder, teams need to be very careful with platoons because of the roster space issue.
fireorlime
7/06
Or they need to change the structure of their roster back to a 10-man pitching staff in order to give them more platooning flexibility?
wonkothesane1
7/09
Pie in the sky.
Pipe dream.
Blow out the candles and make a wish.