Launch Angles Aren’t Enough: Exit Velocity Is King.

Last week I wrote about using pitcher launch angle as tie breaker during a draft. This was an effort to tie the changing home run environment into pitcher value. The correlations are very weak, but it might be valuable on the margins. For example, if you consider two pitchers roughly interchangeable, perhaps you would want to use launch angle to nudge you in one direction or another.

I got feedback asking whether or not launch angle is sticky from year to year, and when it might stabilize. Well, let’s take a look.

We’re looking at the percent of balls in play that are hit above 19 degrees. Roughly 98% of home runs with a Statcast measured launch angle and exit velocity are hit with a launch angle greater than or equal to 19 degrees.  

All of the BIP considered have a recorded exit velocity of at least 20 mph.  This eliminates all batted balls that have gone unrecorded by Statcast, which is about 15% of batted balls.  

There are occasions when fly balls are reported to have negative launch angles, and at least 12 fly ball home runs fall into this category.  I have not corrected for this sort of measurement error for two reasons.  First, the correction is easier said than done.  Second, if you are following along using Baseball Savant, you may not be able to easily correct for this sort of error. As a result, you would be forced to include these errors in whatever calculations you may be doing by hand. By leaving in these errors, the correlations are more representative of your interaction with the stats, even at the sacrifice of being less representative of the actual stats in their truest form.

The analysis is using a simple even/odd split to group batted balls into two separate groups, then finding the correlation between the two, and I am using an r = .50 as the reliability threshold. This is a hotly contested number, you may have read one of the many debates between Tom Tango and Pizza Cutter regarding the merits of using r = .5 and r = .7.  Long story short, when someone tells you a stat has an r = .5 after x plate appearances, that means you can add x number of league average success rates to a player’s total to regress them to the mean.

So, say a stat has an r = .5 after 100 PA, and you have 100 PA worth of data for a given batter.  You can take his 100 PA, add in 100 league average plate appearances, and as a result you’d have the player’s numbers regressed to the mean.

The Numbers

For batters, these flyball launch angles are reliable after 600 balls in play, and for pitchers it only takes 570 balls in play. In other words, a starting pitcher’s stats will be reliable after roughly one full season, while a batter may take a season and a a half. This is longer than I’d like to see, and it really helps to highlight the value of exit velocity, which becomes reliable after about 50-60 balls in play.  It seems that exit velocity is much more important than launch angle, although obviously both are vital.

I tried constraining the launch angles to between 19 and 39 degrees. These angles have extremely high slugging percentage and a very respectable batting average in addition to their high home run rates. Unfortunately, the rate stats did not become reliable after two seasons worth of data. For batters the correlation is r=.168 and for pitchers r=.122. That simply is not good enough to rely upon. It seems that batters and pitchers can control whether a ball is hit into the air pretty well, but hitting it on the ideal angles has much more statistical noise.  Or maybe it requires another variable to help tease out skill, such as exit velocity.

Conclusion

Launch angles are interesting, but only when combined with exit velocity.  Exit velocity is much more important. I contend that the pitchers I identified last week should still be question marks, due to their increased exit velocities.





Andrew Perpetua is the creator of CitiFieldHR.com and xStats.org, and plays around with Statcast data for fun. Follow him on Twitter @AndrewPerpetua.

newest oldest most voted
Gavin
Member
Member
Gavin

“Unfortunately, the rate stats did not become reliable after two seasons worth of data.”

Boo! Hiss!