I’ve written quite a bit this year on trends in pitcher aging, specifically velocity loss and gain. In the last iteration I focused on the odds of pitchers gaining velocity back after a season where their fastball dropped by at least 1 mph.
In that piece I listed a few pitchers to keep your eye on given that their velocity was down from 2011. In June, I wrote about CC Sabathia for ESPN and noted that the big lefty is likely beginning to “age”, as the odds are quite a bit higher that pitchers over the age of 30 do not gain their velocity back once they’ve lost it.
After thinking about it a while it occurred to me that there is of course the chance that these pitchers will gain their velocity back by the end of the year (as I noted in both pieces). We know that, generally speaking, pitchers gain velocity as the season goes on. Temperatures rise, and so too do fastball velocities. If this is the case I wondered at what point in the season we can say with greater certainty that a pitcher is throwing as hard as he is going to throw. Is there a particular month where a velocity decline is more likely to translate to or predict a full season velocity decline?
To get an initial sense of the numbers I looked at fastball velocities for individual pitchers in each month since 2002. I used BIS data as I wanted as large a sample as possible, and BIS fastball velocity data has a correlation of roughly .97 with PITCHf/x velocity data.
I was mainly interested in two numbers: the correlation of fastball velocity in a given month to the rest of season velocity (i.e. how well does April’s velocity predict May through September), and the odds of finishing the season down at least 1 mph when a pitcher’s velocity was down at least 1 mph in a given month from the same time the previous year (i.e. if April 2012’s velocity is at least 1 mph less than April 2011, what are the increased odds of finishing 2012 down at least 1 mph).
I restricted the sample a few different ways. The main thing I wanted to control for was pitchers getting called up late in the season and having fresher arms, or getting injured in May only to return for July and August and biasing the data. Therefore, I set a minimum innings pitched threshold that had to be met in each month of a given season. I also excluded any pitchers that switched roles (i.e. moved from starter to reliever, or vice versa) to avoid any artificial increases or decreases in velocity, year over year.
The first cut I looked at was for starting pitchers that threw at least 25 innings in each month in a given season. Here are the results:
Monthly velocity is generally just as predictive for starters from May through August. There is a little bit of a difference in April, although it’s not drastic (roughly 2%). But the fact that April velocity is less predictive of rest of season suggests that we should take April velocity declines (and, I guess, gains as well) with a grain of salt. This also matches up pretty well with what we find when looking at how the loss of velocity in a month affects the chances of being down at season’s end.
I calculated odds ratios for starters in the sample to see whether losing velocity in certain months was a better predictor of an overall velocity drop than other months. Overall, a velocity loss in any month relative to how a pitcher was throwing the year before increased the chances of a velocity loss for the season, but there are differences by month:
|% with 1 mph drop in month and full season||% without 1 mph drop in month but dropped full season|
What we find is that losing velocity in June or July increases a starter’s odds of finishing the year with an overall velocity loss the most (11.9 and 13.7 respectively for June and July). Losing velocity in April isn’t insignificant, but relative to other months it is less worrisome.
To bottom-line the findings: by June, we should have a pretty good idea of how hard a pitcher is going to throw for the rest of the season. If a pitcher is down at least 1 mph in June relatively to the previous June we shouldn’t expect them to magically recover their velocity in the second half of the season. Part of this, of course, is timing–you don’t have as many pitches left in the season to recover.
In terms of next steps, there are a bunch of ways to improve this study:
1) It would be interesting to look at not just velocity in a given month and it’s correlation to rest of season velocity, but season-to-date velocity through a given month.
2) Monthly data was used because it was the easiest to get my hands on. However, months are somewhat arbitrary endpoints. It might be more telling to look at how velocity trends predict seasonal outcomes based on the number of pitches instead of just months.
3) The odds ratios are good starts, but I didn’t do anything to control for pitchers that were already down in terms of their velocity prior to a given month. So, for example, if a pitcher was down 1 mph in July they may very well have been down 1 in June and even May. So July’s odds may be partially inflated because a pitcher has been losing velocity all year, making it very difficult to make up velocity in the remaining two months of the season after digging themselves such a big hole.
4) Relievers: I did run the numbers for relievers but I am not sure how much I trust them. I used a minimum number of 10 innings pitched in each month as the threshold, which reduced the sample. The correlation between monthly velocity and rest of season velocity looked similar to starters, but the odds ratios in terms of velocity loss looked completely different. There could be a logical reason behind this, but I haven’t thought of one yet outside of the way I selected the sample could have skewed things.
Like many things, this is a first step. Hopefully I’ll get a chance to revisit this or some enterprising readers will jump right in.
In the meantime, here is a “watch-list” of 2012 starters that were down at least 1mph in June and July this year relative to those months in 2011:
|Name||Team||FBv Difference in June (mph)||FBv Difference in July (mph)|
A huge thank you to FanGraph’s intern Sean M. Stouffer for helping me pull the data together for this study.
Bill leads Predictive Modeling and Data Science consulting at Gallup. In his free time, he writes for The Hardball Times, speaks about baseball research and analytics, has consulted for a Major League Baseball team, and has appeared on MLB Network's Clubhouse Confidential as well as several MLB-produced documentaries. He is also the creator of the baseballr package for the R programming language. Along with Jeff Zimmerman, he won the 2013 SABR Analytics Research Award for Contemporary Analysis. Follow him on Twitter @BillPetti.