Platoon splits 2.0

A couple of weeks ago I came up with a classification of pitches, based on speed and movement provided by PITCHf/x, into 14 categories. I’ll list them once more.


type		 speed
Heater 		  96.1   -7.8     6.6
Jumping fastball  92.8   -4.1    10.5
Sinker 		  90.8   -9.2     1.2
Cutter 		  89.8    0.1     6.5
Rider 		  89.8   -9.4     7.0
Rising fastball   86.6   -3.9    10.3

Breaking pitches

type		 speed
Slider 		  84.5    1.7     2.3
Power change 	  84.3   -7.3     4.2
Tight curve 	  81.2    3.1    -3.8
Straight change   79.5   -5.3     6.9
Slurve 		  78.5    5.3     1.3
Roundhouse curve  75.1    6.0    -6.0


type		 speed
Submariners 	  84.2   -6.2    -3.8
Floaters 	  70.5    3.3     2.8

I decided to work in parallel starting from there. One part of me is trying to refine the classifications, the other is exploring various topics starting from that first quick and dirty work (and it’s ready to stop part one in case it finds out there’s no use in spending too much time for that kind of work).

The former has discovered that pitch classification for lefty pitchers seems trickier than the one for righties. It’s also conscious that some more human advice needs to be put into the classification process: The low arm angle pitches, for example, have to be split according to their speed and break, the knuckleballs can’t be classified together with the lollipop curves, and fast change-ups should not be confused with slow fastballs.

Thus, while waiting for the refined classification, the other part of me decided to expand on the works on platoon effects by John Walsh (THT Annual 2008) and Dave Allen (Deconstructing the Non-Fastball Run MapsPlatoon Splits for Three Types of Fastballs).

I have worked only on right-handed pitchers’ data for 2009; the “platoon” column below is simply average Run Value (per 100 pitches) of the pitch against left-handed batters minus average run value against righties. Thus, negative value means reverse platoon split. I left out the special pitches (knuckles, lollipops, submariners).

	   pitch	platoon
          Sinker	 1.07
          Slurve	 1.07
          Heater	 0.79
          Slider	 0.54
          Rider	 0.54
          Cutter	 0.41
Jumping fastball	 0.28
 Rising fastball 	 0.21
    Power change	 0.08
     Tight curve	-0.17
Roundhouse curve	-0.65
 Straight change	-0.77

As both Allen and Walsh showed in the past, it’s the fastballs and the sliders that have the highest platoon splits, with curveballs and change-ups being neutral to batter handedness (sometime even showing a minimal reverse split). Dave Allen also noted that sinkers have the most extreme platoon split, while that of the cutters came out as nonsignificant for him.

Here we see that the jumping and the rising fastballs are the least affected by the batter handedness. Looking at the initial table, they are the ones with the smallest amount of tail (i.e., horizontal movement) toward the right-handed batter.
The same happens with changes of pace, where the power change (the one with significant lateral movement) is platoon neutral, while the slow change shows some reverse platoon effect. Finally the slow curves (drops/Uncle Charlies), having a big arch toward the opposite handed batter, also exhibit an inverse platoon split (stronger than that of the tight deuces).


While reading and commenting the above table, I could continuously hear MGL whispering to my ear “what if sinkers are thrown to right-handed batters in different counts than to lefties?” If you don’t understand what I’m talking about, please have a look at this article.

I turned to an advanced statistical method (multilevel modeling, something I already used in another article) to neutralize the effect of the ball-strike count on Run Value. Having bothered to apply that technique, I thought it could be useful to also neutralize the pitcher effect (just to be sure values don’t get inflated by some very good pitchers throwing a pitch mostly to righties).

Here is the adjusted table.

           pitch        platoon
          Slurve 	 1.12
          Sinker 	 1.08
          Heater 	 0.80
          Slider 	 0.57
          Rider 	 0.56
          Cutter 	 0.41
Jumping fastball 	 0.27
 Rising fastball 	 0.21
    Power change 	 0.01
     Tight curve  	-0.13
Roundhouse curve 	-0.69
 Straight change  	-0.77

As you can see, the numbers don’t differ by much from the previous table, and the ranking of pitches is almost unmodified.

Can this informations be useful on the field?

Let’s say you are writing your lineup against a right-handed opponent. To decide whether having it particularly lefty-loaded, you probably need to look no further than to his traditional split stats. Anyway, knowing that his fastball is a rising one and his change is a slow one (rather than a heater and a power change) might be of some help.

Same thing when deciding which arm to bring in from the bullpen. If you need your reliever to record three outs, you’d probably prefer a heater guy if three right-handed hitters in a row are coming and a jumping fastball guy if it’s righty-lefty-righty.
Hey, but don’t overlook the overall value of your pitchers! If the heater guy is much better than the jumping fastball guy, who cares about the splits?

Homestretch: The 1967 AL Pennant Race, Part 3
A tight race shows no signs of letting up.

References & Resources
John Walsh: The Origin of the Platoon Advantage, appearing in The 2008 Hardball Times Baseball Annual.
Note: You can read the 2008 Annual for free following this link.

Dave Allen: Deconstructing the Non-Fastball Run MapsPlatoon Splits for Three Types of Fastballs

Data from MLBAM’s PITCHf/x.

Print This Post
Sort by:   newest | oldest | most voted

Great article, love reading stuff like this.  But is this not a bit of a waste of time for non-nerds?  Couldn’t I just check my yahoo sports/fangraphs page for splits, whether for one season, a few seasons or career?

Dave Allen
Dave Allen


Very cool.  I really like the way you split up pitches into those finer buckets.