How Does a Different Release Point Affect a Curveball's Effectiveness?

As a Met fan, one of the pitchers I'm very interested in has been the Mets' Jon Niese.  Niese is young, a genuine home-grown prospect, and well isn't one of these super-hyped guys getting all the national attention.  Moreover, as he came up through the minors his big weapon, until last year, was his curveball.  And with that in mind it was impossible not to notice, upon looking at his pitches using pitchFX data, that there is a huge difference in where his release point is for his curveball and where his release point is from his other pitches.  You can see that in the above picture (His curveball is the dark green pitch track, the other pitches are the other colors.) 

Now, the pitchFX system's release point data isn't perfect at calculating release points of pitches; the system essentially backtracks the pitches path to where it should have left the hand.  This results in curveballs in general being recorded as having a higher release point than other pitches (as the large vertical movement downwards causes the system to calculate the pitch to come from a higher angle).  In 2009, of the 187 pitchers who threw at least 100 curveballs and 300 other pitches (using the MLBAM algorithm to classify pitches), only 5 were listed as not having such a "higher" release point...and these 5 all have "curveballs" that have very little vertical break downwards, making me suspect they're not curveballs at all, but rather are the result of a classification error. 

That said, Niese, along with a bunch of other pitchers, have such a large gap between their recorded release points that this gap is almost certainly not just an illusion caused by the PitchFX calculations but a real difference that is observable if you pay attention.  52 Pitchers in 2009 who met my qualifications (100 Curveball/300 Other Pitches) had a curveball release point over 5 inches different from their fastball release point, 32 have a curveball release point that is recorded as being over 6 inches different from the other pitch release point, and one pitcher, Eric Bedard, had a curveball release point that was over a foot different from where he released his other pitches (Niese didn't qualify in 2009, but his release point differential in 2010 is the largest this year at 10.4 inches).

The question thus that has always followed when someone notices this different release point of Niese or one of these pitches is whether batters can notice this release point difference and thus know when the curveball is coming or if for some other reason the release point difference causes a decrease in the performance of that pitch. Is this different release point something that needs to be addressed by the pitcher and his pitching coach? 

The data seems to say the answer is no.

Figure 1 and 2 (Click to enlarge):  Correlation between the farther away a pitcher's curve-ball release point is from a normal release point and the Swinging Strike% and the Whiff% of that pitcher's curveballs. 

To Answer These Questions, I looked at last year's PitchFX data.  I only looked at pitchers who had thrown at least 100 curveballs and at least 300 non-curveball pitches (In order to have enough of a sample size so that swinging strike % and Run Value Above Average data was not completely meaningless.).  That left me with a total of 187 pitchers who threw enough pitches to qualify for this study.  I then plotted the swinging strike % (Swinging Strike % = Total Number of Swinging Strikes/Total Pitches) on curveballs for each of these pitchers against how far apart the curveball release point was.  I also did the same thing using a pitcher's curveball whiff % (Whiff% = Total Number of Swinging Strikes/Total Number of pitches that batters swung at) instead of swinging strike %. 

The above graphs and best fit lines show not much of a prospect for a relationship between having a different release point for the curveball and the performance of that curveball.  The relationship between Swinging Strike % and the distance between the two release points is more or less nonexistent (R² =0.050189) while the relationship with whiff % is even smaller (R² = 0.027311).  So having a different release point does not seem to affect how well a curveball is able to get those valuable swinging strikes.  There's no relationship between the change the release point and the run value of the curveball either (Graph not pictured above, but the R² = .011368). 

It is true however that the above analysis of pitches includes all pitches classified as curveballs by the MLBAM Algorithm and there are of course multiple types of curveballs.  So I next attempted to rerun the analysis limiting the data to only include "12 to 6" curveballs and then to only include curveballs that have a sort-of "11 to 5" motion. 

Figures 3 and 4:  The Swinging Strike Analysis repeated for only "12 to 6" curveballs and "11 to 5 curveballs."

As you can see, these graphs aren't much difference from including all the classified curveballs.  It's worth noting that these are limited sample sizes (especially the 12 to 6 graph, which only includes the pitches of 27 pitchers), but there's no reason to expect anything different with a larger sample size.  I did the same thing for whiff % and had the same lack of result. 

Conclusion: 

Now there is no one factor that results in curveballs (or any other pitch) being particularly effective.  The amount of strikes in the count when the curve-ball is thrown has a particularly strong effect:  the more strikes on a batter, the more likely is the pitch to result in a swinging strike, but this effect is due entirely (as whiff % actually goes down as there are more strikes on a batter) to batters swinging much more often at curveballs when there are more strikes in the count.  There is also a minor correlation (R² =0.165481) that shows that swinging strike % increases when curveballs are more down in the strike zone, particularly when thrown to opposite-handed batters (R² =0.239025).  But in reality, it is a variety of factors (location, the count, what pitches were thrown before hand, the specifics of the batter's tendencies) that determine whether or not a curveball will be successful at getting swinging strikes and whether a curveball will be successful in general. 

What does NOT seem to be much of a factor is whether the pitcher throws a curveball from a clearly different release point.  It's possible that only huge differences in release points (as shown by Jon Niese and Erik Bedard) result in a weaker pitch, but considering Bedard's success (and the low sample size of these extremes), I don't think this is the case.  Thus if a pitcher is seen to be using such a release point, I would not consider it a problem worth looking into; as changing the release point will probably result in little effect and may impact the movement and location-control a pitcher has over that pitch. 

In reality, what this study seems to reveal is that a pitcher doesn't really tip their pitches by using a separate release point for a single pitch (well either that or tipping a pitch has no real effect, which seems very unlikely to me).  This makes sense really...by the time the batter could notice that the release point is different, the pitch is already in the air and he has a tiny amount of time (2 tenths of a second) to process this information and decide whether or not to swing.  And a batter, unlike those of us looking at a pitchFX graph, can't really see as easily whether a pitcher is using a 1 o clock rather than a 12 o clock delivery to make a pitch, as he can't compare the two pitches in real-time. Rather, when someone is considered to be tipping their pitches, they are doing something before they actually throw the ball that gives away to a hitter what is coming. 

Sky's note: Please welcome Josh to the BtB family.  I've only read a few of his articles (check out Amazin Avenue) but I'm quite impressed and look forward to reading more.  This addition should also give us a hollow defense that we don't hate the Mets.