Thursday, June 12, 2008

Penny f/x (originally posted in comments at Dodger Thoughts)

I remember there was a discussion on Penny's pitches in a recent DT thread, and Josh Kalk has just released the 2008 version of his pitch f/x tool ( So I checked Penny's numbers. The average NL pitcher this season has 3.79 pitches per batter faced, with 18.7% of pitches batted, 43.4% of pitches strikes (non-batted), and 37.9% of pitches balls. Penny overall has had 1453 pitches, 1293 of which were tracked. Penny's pitches have been 19.1% batted, 43.3% strikes, and 37.6% balls.

On batted balls, the league has a .347 wOBA against with .207 singles per batted ball and a .187 ISO; Penny's batted balls have had a .351 wOBA, .255 singles/batted ball, and a .137 ISO.

Of the charted pitches, Penny has 67% fastballs, 17% curves, and 16% change-ups.
FB: 37.6% balls, 19.1% batted, .376 battedballwOBA, .259 1b/batted, .162 ISO
CV: 40.6% balls, 16.6% batted, .344 battedballwOBA, .306 1b/batted, .056 ISO
CH: 37.5% balls, 22.6% batted, .267 battedballwOBA, .191 1b/batted, .106 ISO

So the fastball has been hit hard when put into play and the changeup has been pretty effective in generating outs. The [Eric] Enders hypothesis [from a previous DT thread] seems to hold - hitters are either sitting on the fastball or getting very lucky on it (actually, some combination thereof).

Here are the uncharted pitches (11% of his total):
33.8% balls, 18.1% batted, .358 battedballwOBA, .276 1b/batted, .138 ISO

While the fastball has been hit hard, Penny has still been only one run below average on batted balls. What is really concerning is that he's thrown basically a league average split of balls/strikes/batted balls, but has a szERA (strike zone ERA, an ERA estimator using only K, BB, and BFP, although Tango seems to have renamed it kwERA (for K and Walks)) of 4.97 against the league's 4.39. Maybe he just hasn't had the right sequencing, or maybe he just doesn't have the right stuff to put hitters away.

* * *

2007 data is available at the same link. Penny's fastball worked WAY better last season. I had actually already run the 2007 numbers for Penny a few months back.

Total: 35.4% balls, 20.2% batted, .315 battedballwOBA, .218 1b/batted, .117 ISO

Charted pitches: 71.9% fastball, 12.0% curve, 16.1% changeup. 36.9% of pitches were uncharted.

FB: 34.3%, 19.4%, .273, .207, .085
CV: 35.5%, 18.4%, .385, .267, .133
CH: 39.3%, 23.5%, .425, .273, .182
Charted (tot): 35.3%, 19.9%, .314, .227, .108
Uncharted: 35.8%, 20.7%, .315, .203, .130

That's a .102 difference in wOBA on batted fastballs (that were charted)! And the (charted) curves and changes that were put into play were hit harder than his (charted) fastballs have been this season. The difference in the results of his batted pitches has been pretty night and day.

His szERA was 4.54, and part of the dropoff in that can (in addition to variance etc.) be explained by the 2% increase in balls per pitch. And in case anyone was wondering whether his lack of K this season stems from an inordinate number of foul balls, he's actually had a smaller percentage of foul balls in 2008 overall, including a smaller percentage of fastballs.

* * *

A split finger FB should be included among the fastballs. I could be wrong, though. Josh's tool lets you break them down by velocity, and I see that Penny has 69 pitches listed as fastballs that are under 90 mph, so I'm assuming that that includes the splitters. The "uncharted" pitches are ones that weren't captured by the f/x, not ones that don't fit into the classification system. If you're really interested in the splitter and its classification, I think Mike Fast is the guy to talk to since he works with the spin data (and I bet if you asked he'd probably do the research). As I understand it, unless you are looking at spin (the Kalk tool does not) you can't differentiate different types of FB.

* * *

I broke Penny's fastballs in 2008 down by velocity:

<88 6
88 27
89 36
90 45
91 68
92 108
93 128
94 160
95 151
96 91
97 42
98 6

I then broke them down into four groups: a) less than 92 (182 pitches), b) 92+ and less than 94 (236), c) 94+ and less than 96 (311), and 96+ (139).

The six numbers below are average velocity, %balls, %batted, wOBA on batted balls, singles per batted ball, and isolated power.
a) 90.33, 38.5%, 17.0%, .225, .161, .065
b) 93.09, 39.4%, 15.7%, .414, .243, .216
c) 94.99, 34.4%, 21.9%, .343, .309, .074
d) 96.81, 40.3%, 21.6%, .559, .267, .400

Sample size caveats should be peppered throughout any analysis like this, obviously. It would seem that his slowest fastballs (presumably these are mostly split finger) have been very effective. This feeds the "hitters are sitting on (non-splitter) fastballs" hypothesis. The fastballs that have been have been slower have been hammered and also have missed the zone more. The fastballs in his typical range (94-95) appear to have the best command (by ball%) and have not been hit hard power-wise (.074 ISO) but have been slapped for a ton of singles. The fastest fastballs have been hammered when hit (both in terms of power and singles) and have missed the zone quite a bit. If you check out the plot (, he just does not seem to have good command on the fastest fastballs (though a) sample size and b) I don't know the general spread): the hits have all been out over the middle of the plate and, except for two that were kept down, have been in the upper half of the zone without being in the upper sixth or so of the strike zone.

Taking a look at the strikes that weren't put into play, here are the percentages of called strikes, swinging strikes, and fouls in that order:
a) 51.9, 9.9, 38.3
b) 45.3, 9.4, 45.3
c) 25.7, 15.4, 58.8
d) 30.2, 13.2, 56.6

Including ALL pitches (not just strikes), the swings per pitch are a) .385, b) .403, c) .543, d) .482. So it seems that hitters have laid off of or have been deceived by the slower fastballs (again, hitters that can differentiate them seem to be laying off the splitter, bolstering its success) and batters are not having too much difficulty getting the bat on the higher velocity FB's. Balls in play per strike shows a) .306, b) .279, c) .362, and d) .311, so it really does seem that hitters are waiting for a typical Penny fastball and slapping them for singles.

So the data seems to back up Eric's hypothesis very well - well done!

UPDATE: I've been trying to learn more about Penny's repertoire and whether it's as simple as FB/CU/CH or if there are (as I have seen suggested from various sources in this search) splitters, sinkers, sliders, or 2-seamers (or, as some state, no true change-ups). I really don't think I'm any closer. I've looked at a lot of this. And this and this. I wonder if we could get Aaron Hintz and Harry Pavlidis to come to a consensus.


Post a Comment

<< Home