Thanks to Matt for filling in with some interesting explorations on spin rates the last two weeks. By the way, let me know (email me at firstname.lastname@example.org) if you would like to contribute to the blog with some R baseball exploration — I appreciate the breaks!
Platoon Advantage and Platoon Effects
In my last post, I explored platoon advantage, that is the percentage of times a batter faces a pitcher of the opposite arm. In that post, I talked about an interesting issue when one looks at platoon effects, that is how much better one performs against a pitcher of the opposite arm. The problem is that players with perceived large platoon effects will likely be platooned and they won’t get many opportunities to face pitchers of the same side. So if you do an analysis using players with, say at least 100 PA against each arm, you are eliminating some players with large perceived platoon effects.
To illustrate this issue, let’s look at Ryan Howard, who (I believe) ended his career with the Phillies in the 2016 season. (I have a display of all of the baseball cards of the 2008 World Champion Phillies and Ryan was the last one left on the display who was currently a Phillies player.)
How Much Was Ryan Platooned?
In Ryan’s early seasons, he was a star playing about everyday, so I believe his platoon advantage would be relatively small. But as he aged and declined in hitting performance, things changed. Using Retrosheet data, I graph the fraction of PA’s against pitchers of the opposite arm against season. We see that he was platooned a lot early and late in his career. In his best seasons (2007-2008), his platoon fraction was a little over 60%. In contrast, he was platooned about 90% of the time in 2016.
Ryan’s Platoon Effect
Okay, one would think that Ryan is being platooned since he wasn’t hitting well against left-handed pitchers. Using wOBA is our measure of performance, I graph Ryan’s wOBA against lefties and righties below. The size of the plotting point corresponds to the number of PA’s — this gives you an idea of the platoon advantage.
We know Ryan’s hitting performance showed a steady decline starting with the 2010 season. But what is interesting is that the size of his platoon effect (that is, how much better he performs against pitchers of the opposite arm) appears to be pretty constant in this period of decline. Actually, Ryan’s performance against righties in 2016 was actually lower than his performance against lefties in the early part of his career. Notice the small red dots indicating the small number of PA’s against pitchers of the same arm in the later seasons.
Comparison with Similar Players
Next I wondered if Ryan’s platoon advantage (and decline in overall performance) was similar to that of similar players. After looking at Ryan Howard’s Baseball Reference page and some thought, I came up with Jim Thome, Prince Fielder, and Jason Giambi. All of these players are left-handed sluggers who played about the same time as Howard.
Again, using Retrosheet data, I first looked at the platoon advantage for each of these players. Giambi, Thome, and Howard were all platooned more in their advanced years. Prince Fielder was only platooned heavily in his first season. Generally these patterns of platooning are pretty similar.
What about their platoon performances? We’ve already talked about Howard’s large platoon advantage. Giambi, in contrast, did very similar against pitchers of both arms. Thome exhibited a large platoon effect, but I don’t see the decline in overall wOBA performance as severe as that for Howard. Fielder also had a platoon effect, but not as large as that for Howard and Thome.
I think the platoon effect issue deserves further exploration. It seems that most managers believe two things — first, they believe the platoon effect is large which means that they will regularly platoon players to give their team a supposed advantage. Second, they believe that players tend to have the same platoon advantage. I think instead that platoon advantages differ between players, and I would be interested in finding the type of hitters who have modest platoon splits.
Data and R Notes
These graphs were relatively easy to produce from Retrosheet play-by-play data — the key variables in the data frame are BAT_ID (code for the hitter), BAT_HAND_CD (batter side), and PIT_HAND_CD (pitcher side). I have a single data frame that I used with all of the Retrosheet data from seasons 2000 through 2016, over 3 million rows. I used data from a spreadsheet from FanGraphs that provides the weights for computing the wOBA values.