How Many Pitches Does it Take? Part One
I’ve been talking about pitches, pitching patterns, and pitch usage a lot lately. Whether it be through PitchFx charts, simply sharing observations, or talking about a pitcher who needs an additional pitch. Finally, I broke down and gathered the data needed to see whether having a surplus of pitches or only a couple mattered to performance.
Most people have the idea that quality matters more than quantity in mind. I know I did. In fact, while running the query (last three years, at least 5% usage of the pitch, at least 150 innings) only one pitcher recorded more than five pitches and that was Ryan Franklin with six. As you’ll see, Ryan Franklin is not a particularly good pitcher. Franklin is passable, but I think you would expect more from someone who has a constant advantage in game theory. Now, it is possible that Franklin falls into patterns, tips his pitches, or simply throws hittable garbage, I’ll leave that up to you to figure out, my only interest is the amount of pitches used modestly and whether it makes for better pitchers.
We begin today with that query I mentioned earlier. No restriction on amount of games started and only 150 innings over the last three years; meaning relievers like Joe Nathan, Mariano Rivera, and Jonathan Papelbon were eligible to make the cut. Let’s get to the data, shall we?
Franklin was the only pitcher with six pitches and 28 pitchers had five pitches qualify. Tradition has most starters throwing 3-4 pitches and most relievers having one or two. Tradition holds true here. 119 pitchers had four pitches qualify, 134 had three, and 39 had two. Franklin failed to make a start over the last three years meanwhile pitchers with 5 pitches saw 64% of their games come as starters, 60.2% for four pitch qualifiers, 32.2% for three pitches, and 11.7% for two pitches.
Let’s look at how they actually performed:
Are the relievers skewing the two and three pitch numbers? Tomorrow we’ll separate the starters from the pack and see if that’s the case.
This is an interesting line of analysis. I’d never thought about actually trying to see whether more pitches helps. Still, though, it seems like the actual value of having more pitches is going to be dwarfed by the ‘junk-ball effect’–the pitchers who don’t have enough stuff or speed to make it on their first 2-3 pitches are going to work harder on developing different pitches.
Maybe you could focus on individual pitchers’ difference in performance in the year or two after they develop a new pitch?
Ooh, good idea.
I’ll see if I can’t find a few after I wrap this series, but did you have a certain pitcher(s?) in mind?
Well, as a Mariners fan, the example that comes to mind is JJ Putz–there was a clear example of how adding another pitch helped in a game theory sense. He added a splitter that looked similar to his fastball from the delivery. It helped that it was a very good splitter, of course.
Is there a way that you can systematically look at pitchers who add pitches from one year to the next, or can you only look up pitchers on a career basis?
Halladay (cutter), ErvSantana (slider) would be two I’d like to see if you have the time.
How about White Sox pitchers who were taught the cutter after joining the organization? Jenks, Contreras, Danks and Esteban Loiaza come to mind — as do guys like Aardsma, Sisco, MacDougal, etc. It’s been a fairly well-documented strategy from the ChiSox’s scouting side, but I’d be curious to see if it truly pays off using this form of analysis. Just my two cents …
Lincecum’s changeup may have been a major help
I’d be interested to the distributions in this chart by team. I know that Toronto seems to encourage its pitchers to throw lots of different pitches in games; do other teams have easily identifiable tendencies?
I think the results you see are self-selection bias. The more marginal a pitcher is then the more pitches he try to throw. The better a pitcher is then he needs fewer pitches to be effective.
So the conclusion is definitely not that Ryan Franklin needs to mothball 3 of his pitches to get better results.
And wow, I didn’t realize FIP correlated so well with ERA when aggregating pitchers.
Are you distinguishing between fastballs? e.g. are 2 seamers and 4 seamers and cutters considered distinct pitches?
Pitch Types that BIS tracks:
FB – fastball
SL – slider
CT – cutter
CB – curveball
CH – changeup
SF – split-fingered
KN – knuckleball
I’m sure there are some classification errors in terms of fastballs, slurves (is it a slider or curve?), etc. though.
I know it would be very hard to track, but there is a significant difference between a 2 seam fastball and a 4 seam fastball. I always threw exclusively a 2 seamer until I couldnt locate it at all (until may 31st).
Then, for the first time in my professional career, i threw a four seamer (instead of a two seamer) to get better command of the strike zone (ie 0-1 vs 1-0) and had tremendous results. The drastic differences in my stats (before and after may 31st) were due basically to me throwing a 4 seamer and not a 2 seamer which allowed me to throw more strikes.
But other than that anecdote, this is pretty cool analysis. I would like to see the stats broken down by SP and RP.
What about someone like el duque, who throws about 8 different “pitches” due to arm angle and speed, even though many may be classified as the same thing?
For now stuff like arm angles, grips, etc. is not being classified as different pitches.
I think Kazinski is likely on to something, though it will be interesting to see what happens with the relievers removed. The lower the quality of your offerings, the more need you have to deceive the hitter through other pitch types. If you’re Mariano Rivera and can throw an unhittable cutter, why develop a curve — or at least use it.
Less pitch variety = more dominant pitches = higher k/9, lower HR/9, and lower FIP, but leads to overthrowing = higher BB/9 (usually).
BABIP is flat meaning a hitters BABIP is a true measure of their own skill and independent of pitching?
Great stuff
Two things instantly pop-up to make this largely a futile exercise:
– Sample sizes issues. Only really have sample sizes large enough for the pitchers with 3/4 pitches to draw some real conclusions. I am willing to bet the 95% CI intervals are pretty large for the other sample sizes and makes it pretty dubious to draw any conclusions from them with much certainty.
– pitch f/x data is interesting but as several others have pointed out it still does a fairly inaccurate job of classifying a pitchers’ true repertoire because of the number of pitchers who throw a modified fastball or a offspeed pitch that can be difficult to classify.
– Given that you have larger sample sizes for the 3/4 pitch groups, I would be curious to see breakouts between starters who are primarily fastball/sliders guys (norm usually today) vs. fastball/other type of breaking pitch