As a facility predicated on data-driven player development from the beginning, we’ve been fortunate enough to collect data on a wide-variety of athletes ranging in age, skill-level, and programming type. Rather than silo this data internally, we annually publish the results for the majority our summer pitchers and open source the underlying raw data as well.
We hope this approach has allowed coaches, parents, and athletes to draw their own conclusions on the merits of our throwing program, which we feel is far more conducive to an open and honest relationship with the baseball community.
We mention this because in this article, we report on the training outcomes for our Summer Hitting Trainees for the very first-time through the lens of exit velocity (EV) on balls in play (BIP).
Why Exit Velocity?
EV is the preferred metric for evaluating the talent levels of our hitters, because it objectively measures how fast the ball leaves an athlete’s bat. This allows us to filter out any subjective measurements of “hard hits” and/or batted-ball luck contained in “x” statistics such as xwOBA.
Beyond these reasons, EV is also preferred for the following reasons:
- Batted balls that are hit the hardest have the most optimal outcomes, generally speaking.
- EV is more reliable than other BIP metrics (meaning that it is likely to be more reflective of the talent level of the batter, holding sample size equal).
- EV allows us, when looking at the hardest hit BIP by specific batters, to gain a proxy for peak bat speed or the change in peak bat speed over time.
- Lastly, when EV unexpectedly increases for a given player, he tends to outperform his expected outcomes at the plate over a given season, which will be explained later on.
Limitations of Exit Velocity
While we believe EV is the best metric for evaluating the development of our hitters, for the reasons above, it also comes with several limitations that make it a bit more nuanced and difficult to report on. Consider the following:
- While hitting the ball harder is almost exclusively a net positive in terms of outcome, gaining EV will have a relatively minimal impact for batters who have an attack angle oriented towards the ground.
- EV is less reliable than pitching velocity, meaning that we need a much larger sample to gain a better idea of how a batter improves over time. (Two at-bats worth of pitching velocity contains about as much information as 225 BIPs of EV.)
- Unlike pitching velocity, which is mostly a function of the pitcher’s ability to throw hard, EV is at least somewhat reliant on the difficulty of the pitch to hit and the context of the pitch itself.
- Since EV is more difficult to calculate than pitching velocity, and swings usually occur in quicker succession, more sophisticated collection tools and methods are needed to accumulate a reasonable amount of player-specific data.
To overcome some these hurdles with regards to EV, we’ve made large-scale additions to our hitting assessment, incorporated more “test days” for athletes, and integrated third-party technology into TRAQ in order to seamlessly accrue batted-ball data on a daily basis.
As a result, we were able to track batted balls in a variety of different contexts for over 50 athletes during the summer months. The tables below illustrate the net change in both average and peak EV for our summer athletes (top table includes all athletes; bottom table excludes any athlete staying for less than three weeks) between their initial and final test days at the facility.
A couple things to note when looking at the graphics above:
- All tests/re-tests are done during front-toss, not against a hi-speed pitching machine.
- Switch hitters are separated out by handedness to avoid comparing left- vs. right-handed swings.
- Hitters are tested using both their game bats as well as our Axe Bat weighted bat speed system.
- We prefer to report on peak EV (defined here as the hardest hit batted ball) over average EV because it is a more reliable metric; it most likely better represents the underlying talent level of a given hitter in smaller samples. Basically, a few unlucky mishits can significantly bias average EV downwards when BIP numbers are limited.
- With that said, we also want to help batters avoid mishits whenever possible. So, we find it important to report on both peak and average EV in regard to game-bat performance.
As shown above, batters improved across the board in EV by 1.5-3.5 mph, regardless of bat type. The largest gains occurred when batters swung their game bats, which is what we primarily try to optimize for to encourage transfer to in-game swings.
In observing peak EV amongst all bat types, batters produced the same peak EV when comparing their game bats with Axe’s End-Loaded and Handle-Loaded bats. However, batters who swung an Underload Bat produced an average of roughly +5 mph on their peak EV when compared to any other bat type.
Beyond just looking at raw averages of EV, the standard deviation of both game-bat-average EV and game-bat-peak EV decreased from initial to exit test, particularly for batters who stayed for three weeks or longer. This indicates that long-term stay athletes were making more consistent contact during re-tests.
Lastly, the average batter who stayed longer than three weeks gained an extra ~1 mph in peak EV when compared to batters who stayed for two weeks or less.
Turning to Results Against the High-Speed Pitching Machine
Beyond just collecting EV data during re-tests, our athletes also have BIP results generated and collected against our high-speed pitching machine. By pulling and analyzing this data, we’re able to apply a more realistic context to any analysis we choose to run on our hitters.
For the scope of this article, we selected BIP data for sessions that could be matched to a playerID from May 1 to August 1. We then calculated the net change in both average EV and peak EV for each athlete with at least four sessions recorded on-site.
Because there is a significant acclimation process for athletes to adjust to our high-speed pitching machine (as shown in the graph below and to the right) and daily performance can fluctuate based on the pitch-type settings of the machine within a given session, we used the following selection process (provided in the table below and to the left) to create estimates for pre- and post-talent levels for hitters training at Driveline.
Once the average EV and peak EV metrics (peak EV is calculated using the 1/8th rule) were calculated for each athlete pre and post training, we were able to find the net improvement of our hitters under a more realistic context.
Based on the numbers above, our summer 2018 athletes gained roughly +6 mph of average EV and about +2.5 mph of peak EV against our pitching machine. Given that the “all athletes” sample, including those who were only in gym for less than three weeks, gained a greater amount of average EV than athletes on-site for three or more weeks, we can reasonably hypothesize that we picked up some beginner gains in athletes who only accumulated data for five sessions or less. However, given that athletes here for more than three weeks gained a similar amount of average EV and a greater amount of peak EV, those beginner gains are likely minimized and do not significantly influence the results.
To visualize the distribution of these changes in EV, we bucketed and smoothed all athlete results by .5 mph increments of peak EV. We also generated a density plot comparing gains in peak and average EV in the space below.
In looking at the distribution table above, we see that 79.43% of our summer athletes reported an increase of at least .5 mph of peak EV in our pre/post peak EV Metric while training on-site, 12.06% of athletes held roughly the same pre/post peak EV metric (within +/- .5 mph) while training on-site, and 8.51% reported a decrease from pre/post peak EV while training on-site.
What do these changes in performance for our athletes equate to on the playing field?
It’s hard to say the exact degree that these gains are fully transferable to in-game performance, given that the athletes within our sample were not facing live pitching due to limitations in sample size. But, given that in-gym athletes face game-like velocity and a variety of different pitches when inside the cage, we believe our BIP metrics are fairly representative of in-game performance. This is evident when you compare the distributions of average and peak EV metrics we’ve obtain against our machine with the expected in-game numbers of our athletes.
While transfer of training varies from athlete to athlete, we feel strongly that our program is designed to promote significant carryover from the performance gains in our facility to performance outcomes on the field.
What is the value in increasing my EV at the plate?
While several studies have investigated the positive relationship between pitching velocity and performance, many have overlooked the potential net benefit of increasing EV at the plate for a given batter.
To study this a bit more closely, we grabbed Steamer Projection Data for batters during the Statcast era and calculated the net change in both average and peak EV from year to year. In joining these two datasets, we found that batters who outperformed their average EV from the season prior by 1 mph also outperformed their expected wOBA by ~7 points as well.
As shown above, the relationship in improving EV and performance is linear and consistent over multiple years. (To be precise, the simple linear model on the left predicts a 6.7-point change in wOBA for a one-unit change in EV, whereas the “poor man’s regression” on the right predicts a 7.3-point change in wOBA for a one-unit change in EV. The r^2 values are .112 and .8123 respectively. Both are drawn off a sample of 928 MLB players from 2016-2018 who recorded at least 100 BIPs in consecutive years and had a coinciding Steamer Projection.) Of course, one might say that it is obvious that hitting the ball harder on every BIP is going to result in an increase of expected performance, so we repeated this exercise only looking at peak EV (or 1/8th of a player’s BIP) and found the same relationship.
In comparing this to the pitching side of things, we know that 1 additional mph of velocity subtracts ~.25 runs off of a player’s RA9 and that 1 additional mph of EV increases a player’s expected contribution at the dish by about 7 points of wOBA (or about .006 runs per PA). As a result, we can multiply these two numbers by the amount of expected IP / PAs for a respective starter on the mound and in the field to compare the values against one another.
By assuming a season’s worth of IP (150) and PAs (600), we find that a +1 mph gain in EV is worth about ~85% of a +1 mph gain in pitching velocity. As a result, it’s clear that improving either peak or average EV by a non-zero amount can facilitate a significant boost in player production.
Moving forward, we hope to continue achieving positive results for both our on-site and remote athletes. Hitting may have fell behind pitching from a data-driven, developmental perspective, but that doesn’t mean there isn’t time to catch up.
As always, the raw data that accompanies the analysis is available.
MLB exit velocity data accessed via baseballsavant.com
– Exit velocity graphics use 2018 MLB Statcast data, which includes 114,268 BIP.
– Reliability graphics used 2015-2018 MLB Statcast data for all batters who have had at least 500 BBE during the Statcast era, giving us a sample of 356 batters.
Steamer Projections courtesy of steamerprojections.com.
wOBA values courtesy of fangraphs.com.
Write-up and analysis for this article was done by Dan Aucoin, the data was mined by Alex Caravan