(Photo by Scott Winters/Icon Sportswire)
In the depths of Statcast are a handful of statistics that are tracked, but not displayed prominently. A few such are pitch acceleration metrics, measured at the point of release in three directions. These statistics measure how a pitch should move in the air due to the forces of drag, Magnus effect, and gravity.
A pitch with high acceleration in air intuitively should be harder to hit than one that is flat. To calculate the sum of acceleration of a pitch in three dimensions, it can be represented as the magnitude of a 3D vector (√ax2 + ay2 + az2). As results tend to cluster around the same baselines for average acceleration on each pitch type, scores for each pitch are represented by its percentile among other pitches of the same type. Pitchers are given an overall score of the average percentile of each pitch they throw.
Firstly, it is important to establish that acceleration percentiles (ACC%) are repeatable year to year to show that pitchers have influence over the effects of acceleration in air. The samples used were pitchers who threw at least 50 innings between 2016-17 (227 pitchers) and 2017-18 (120), as Statcast data past 2015 is more reliable.
| Year | r2 of ACC% with itself year+1 |
| 2016 | 0.516 |
| 2017 | 0.791 |
In the same sample, these correlations compare similarly to groundball rate allowed and better than strikeout and walk rates, all three well established repeatable skills. The 2017-18 r2 was likely higher because almost all pitchers to throw at least 50 innings in 2017 and 2018 so far were starting pitchers who had higher innings totals to stabilize. ACC% also shows some predictive value in pitcher performance on batted balls:
| Statistic | 2016 r2 with wOBA allowed on batted balls year+1 | 2017 r2 with wOBA allowed on batted balls year+1 |
| ACC% | 0.0264 | 0.028 |
| GB% | 0.0187 | 0.025 |
| HR% | 0.0176 | 0.01 |
In both time frames, ACC% performed slightly better than GB% in predicting pitcher wOBA on batted balls. Pitchers with high acceleration percentiles tended to allow weaker contact. While still not the strongest correlation, batted ball results are the most confounding element of pitching to project. ACC% should be used in conjunction with GB% in understanding how a pitcher may have the ability to suppress contact.
I tested other relationships between ACC% and several statistics such as K% and BABIP and found some other modest correlations, but the most unique value compared to existing statistics came in evaluating batted ball profiles. Another possible application could be noting that a sharp drop in ACC% may indicate a pitcher is injured or has a mechanical issue.
The following are the 2018 leaders among qualified SP and RP. Of course there are some good pitchers who succeed with low acceleration and some mediocre pitchers who remain middling despite high acceleration. ACC% still provides a unique insight to pitcher ability that appears to be at least equal to and possibly more than the value that GB% provides.
Chris Sale’s score of 0.96 means that his average pitch is in the 96th percentile of acceleration adjusted for type. While 0.5 is average, starting pitchers tend to score slightly lower than relief pitchers.
SP Leaders
RP Leaders
