Continuing down the unbeaten path of draft models; the last version of this was less than a week ago when I discussed a tricky metric in vertical approach angle.
This time I’ll be going over a rather broad pool of data. Batted ball profiles can be as simple as exit velocity and launch angle, or you can branch off and turn those numbers into much more.
Many teams actually do use data like EV/LA to produce neutral-site expected numbers. And some teams even take it a step further to take numbers that have actually happened and factor in ballpark adjustments, as well as opposing pitching quality (aka stuff+ against). Doing this puts every player on an equal playing field with the right weights/coefficients and helps teams with identifying outliers to target and look more into; because at the end of the day, batted ball profiles are a small but important piece of a larger puzzle.
With specific weighting-based data available, teams can accomplish all of this and then some. We’re going to try and replicate what data teams are looking at. For this article, I’ll be going over numerous offensive metrics and include real life data from one prospect so we can get a glimpse into what MLB teams are also looking at.
We’ll be using the fifth overall pick in the 2020 MLB Draft, Austin Martin, as an example. Now there are a few caveats to note. All of the data is from 2020 only, and because Martin’s first three games (Michigan, Connecticut, and Cal Poly) don’t have any trackman data available, we’ll be ignoring those three games.
As boring as it seems, using just exit velocity and launch angles is perfectly normal. The exit velocity of a batted ball event (BBE) is the biggest variable in terms of predicting BA and SLG. And the difference between EV and LA is quite big for both, shown in the image below:
We know SLG% is one of the best non-weighted statistics (shown in the charts below) when it comes to projecting runs per game. Far better than BA and slightly better than OBP. So, we should value exit velocity far more than launch angle – but having both (when I say both, for launch angle I mean tight distribution) is obviously beneficial for the team evaluating him as its less work in player development.
Now let’s start looking at Austin Martin. Here are his 35 batted balls available through trackman:
It’s a pretty self-explanatory graph. This is exactly what you’d expect from the widely-regarded best pure hitter in the entire 2020 draft class. What makes it so good? well, 26 of his BBE were higher than the average exit velocity in college baseball last year (that being 87.18 in the data I have), 13 were in the triple-digits, 23 of his LA’s fell between the ideal 0-25 degrees range. And lastly, 14 of the 35 were greater than 90 mph and between 0 and 25 degrees.
Because of college hitters using metal bats, exit velocities don’t entirely translate (college data, but Cape Cod data would) and on average, EV’s drop about 2.5-3.2 mph in a full seasons long worth of data in the minors. Here are the averages in MLB last year for comparison:
Comparing Martin to the rest of college baseball with the data I have, his average EV of 94.1 ranks 22nd (between hitters w/ 20 or more BBE) in all of D1 baseball in 2020. Assuming his 94.1 average EV drops off 2.5-3.2 mph like I said earlier, he’d be above average in every category above. When you separate his EV’s, you get an average EV of 95.7 against LHP (6 BBE), and 93.8 (29 BBE) against RHP. Both would still be above average in the MLB; once again, thats assuming they only drop off about 3 mph
Moving on to a more interesting aspect of batted ball profiles, some of the stuff below is a work-in-progress like all public models nowadays and we’re always looking to improve the residuals to make it more predictable.
I talked about neutralizing data and putting all players on an equal playing field. So let’s look at expected BA and SLG. Huge shoutout to Chase Seibold, an intern at BaseballCloud that put a lot of work into this. These are coefficients for the formula, this is the more simple version for xSLG (left), and somewhat simple version of BABIP/xBA (right).
Here is the more complicated version of xSLG, but it’s far better predictability wise.
Now back to Austin Martin. We plugged in all 35 BBE and it spit out these numbers:
|xBA||xSLG||xBA+SLG||Real BA||Real SLG||Real BA+SLG|
And now lets see some video. Here’s Martin’s highest xBA:
This BBE had a 105.2 EV and 8.4 LA which produced a .443 xBA and 1.002 xSLG, both the highest for each metric. The next closest xBA was .442, followed by two at .426 and one at .425. For xSLG, the next closest to 1.002 were .981, .978, and .954.
Here’s a look at a home run against St. Louis:
This one had an xBA of .346 and xSLG of .918. Bit surprising to see the xSLG not surpass 1.000, as with a LA of 29.68 it should be a bit higher, and with a 100.49 mph EV, it’s a ridiculously good BBE. But it also would’ve likely been a fly out if hit to left/center-right/center field as it traveled 388.9 feet.
Similarly to vertical approach angle, batted ball profiles are just a small part of scouting. Given how little data is useful on the offensive side (most of the ‘useful’ stuff is blast metrics and swing path scores) it’s a far bigger piece of the pie and some teams probably find themselves taking players later on in drafts solely based off their elite exit velocities or launch angle distribution.
While I only talked about Austin Martin, it was less about him and more about using an at-the-time amateur hitter who made perfect sense to use as an example for the data we talked about.
This won’t be the last version of this, and there are most definitely more important metrics than exit velocity and launch angle; so we’ll be going over some of those later on down the road.