Adjusting Exit Velocity For Pitch Speed And Location

The relationship between batter exit velocity and league exit velocity is not fully understood. There are many factors to exit velocity that a batter controls. Some of these are physical, such as bat speed and swing path. Others are more psychological, such as pitch selection. However, the batter certainly does not control everything. Pitch speed, for example, is a big factor. There are also environmental effects, like temperature and humidity.

I have been working on this problem with Eno Saris for some time now, bouncing ideas, building small projects, and examining the results. Some of it has been fruitful, others have fallen flat, but each time I feel like I’m getting closer to an answer, and along the way I have accidentally bumped into useful nuggets. Today I want to share one of those nuggets with you. I call it adjusted Exit Velocity, and it is the result of combining and comparing batter exit velocity, league exit velocity, pitch location, and pitch speed.

Yesterday, Eno Saris wrote a bit about our findings, which I suggest reading. Today I wish to explain the methodology, delve a bit into the findings, and conclude with how it may be useful to you going forward.

The methodology.

I have split the strike zone into a very large number of buckets, using both the x and z coordinates (in baseball, the y coordinate points from home plate to the pitcher) and the pitch velocity. In other words, we’re looking at a three dimensional space where the coordinates consist of the x and z location along with pitch velocity.

I split this space up into lots of small chunks, 2 by 2 inches, 1 by 1 inch, 6 by 6 inches, etc. I split it up a total of 17 different ways, creating hundreds of thousands of ‘buckets’, which I weighted according to their relative size, and averaged together. So, the 1 by 1 inch portion is weighted more heavily than the 6 by 6 inch portion, for example. I used these weighted averages to calculate league average results, which I then compared to the batter’s actual game results. 

This is a naive approach, I know there are much more sophisticated methods that could be used here, but this should paint a decent enough picture for what is going on. In the future, more sophisticated methods can be used to narrow it down further. I’m using a sledge hammer here, not a scalpel. But, hey, sledgehammers have their place.

At this point, naming convention became a problem. And I am terrible at naming things. So I called the batter’s exit velocity EV. That’s easy enough.  Next, I have two different types of league exit velocity. First, the actual average exit velocity for the whole league. Second, the exit velocity the league produced with pitches similar to what the batter faced. The former I call lgavgEV and the latter I call lgEV.  Maybe (probably) you can come up with a different way to label two different types of league average exit velocities.

Anyways, once the calculations are completed and the naming convention is successfully hurdled it is a simple manner of averaging all the data for each batter.  Their average EV, and average lgEV. Easy enough. Next, subtract the lgEV from the lgAVGEV (I’m so sorry) to find the adjustment (adj) and then subtract that number from the EV to find the adjEV.

I hope that makes sense.  Here, let me type it out in a formula:

EV – (lgavgEV* – lgEV**) = adjEV

*lgavgEV: average exit velocity for all batted balls

** lgEV: average exit velocity for pitches similar to what the batter faced

This method was designed under the assumption that the adjEV would be the interesting aspect. However, that did not turn out to be the case. Instead, the adjustment itself is interesting. In the formula above, this part:

(lgavgEV* – lgEV**) = adj

The results were unexpected.

As Eno pointed out in his article yesterday, the results were unexpected. I expected to find some sort of measure of bat plane or perhaps bat speed. Instead, this appears to highlight plate discipline.

Although, perhaps this shouldn’t have been a surprise. As I highlighted two weeks ago, pitches in the strike zone have higher exit velocity and higher wOBA/xOBA. That wasn’t surprising, nor should it be. So, with this adjEV, we are sort of indirectly measuring how often a batter makes contact with balls that are in the strike zone.

Of course, certain batters may have hot or cold zones in certain areas of the strike zone, for a large number of reasons. Maybe a given batter cannot hit a high fastball, so pitchers throw him a lot of high fastballs. In that case, his exit velocity in the top of the zone would be depressed. Likewise, maybe they are great at hitting inside pitches, or low pitches, or whatever it might be, which would increase their exit velocity relative to the league.

I assumed these hot and cold zones would be the driving force here.  They aren’t. Location and velocity of the pitch seems to matter more, on the aggregate, and as a result this stat appears to highlight the batter’s ability to choose the correct pitches to attack, and which to lay off.

Caveats and things to keep in mind.

Two years ago I saw an argument for the virtues of the swing and miss. In essence, it argued that good hitters will swing and miss at bad pitches, and bad hitters will hit the bad pitch. This felt counter intuitive at the time, but it has since changed my perspective on batting quite a bit in the years since.

This concept of a beneficial swing and miss might come into play with this adjEV. I am only accounting for balls put into play, so called balls/strikes, swing and misses, and foul balls are all ignored. At some point I will include those factors as well, but at the same time I know it will greatly increase the compute time, which is why I haven’t tried it yet.

Next, batters know their swing better than we do. These players aren’t stupid, and they strive to be the best they can possibly be. They have game plans and strategies that come from years of experience. They know, at least to some degree, which pitches they can handle and which they cannot. Yes, certain batters are better at this skill than others, but all batters have this ability. If they didn’t, they wouldn’t be in the major leagues.

This knowledge will change their approach at the plate, at least to some degree. A batter with a given swing may be able to hit pitches that a batter with a different swing cannot handle. These pitches might be within the strike zone, or they might be outside of the strike zone. This is something to keep in mind. Although it probably has a small impact on the overall numbers.

Finally, pitch location data in 2017 isn’t as reliable as it was prior to 2017. It is just something we have to live with.

Batters who changed between 2016 and 2017.

In Eno’s post you can find data for 2015-2017 adjEV and 2017 adjEV for all of the batters. Today I’d like to show you the batters who changed the most between 2016 and 2017.

In the final column, which I claim measures the change in plate discipline between seasons, positive numbers represent more discipline and a negative number represents less discipline. So, Lowrie, Baez, and Davis are each more disciplined in 2017 than they were in 2016, while Suzuki, Perez, and Grandal are less disciplined.  Note, this third column is relative to the 2016 season.

The second column shows the discipline for 2016, and the third column shows 2017. In both cases, positive numbers show more discipline. So, for example, Jed Lowrie went from mildly undisciplined to well disciplined, according to this stat. Meanwhile, Grandal did the opposite, going from well disciplined to slightly undisciplined.

Overall, the discipline between 2016 and 2017 has a correlation of .69 for all batters with at least 50 batted balls. It jumps to .73 for batters with at least 150 batted balls. So, it is reasonably stable from year to year, but not exceptionally so.

I decided to take a look at Javier Baez, so I imported his heatmap into a tableau viz, and you can see the results below. First, take a look at his 2016 map.

The strike zone, the black rectangle, is his strike zone as reported by pitchfx and statcast, using the sz_top and sz_bot stats you can find in gameday or baseball savant. Notice how far out of the strike zone Baez was willing to swing in order to create contact. Notice how hot the sectors are out of the zone when compared to within the zone. Much of the actual strike zone is light orange and blue. This is far from ideal. Now look at his 2017 heat map.

Yes, Baez is still reaching out of the zone, but to a far lesser extent. The area to the bottom and right of the heat map that were full in 2016 are now largely empty, and the top of his strike zone, a weakness last year, is now red with higher exit velocity hits. Javier Baez has made less contact with out of zone pitches, and better contact on pitches in the zone.

Previously, I mentioned an argument I once heard regarding the benefits of the swing and miss. Baez appears to be a beneficiary of this phenomenon. In 2017, he has a higher out of zone swing rate, but a much, much reduced out of zone contact rate. Meanwhile, his in zone swing rate is up, but in zone contact rate is down.

In other words, Baez is swinging more at everything, and hitting less of everything. Is that a good thing? When you state it so bluntly, it sounds bad, but then when you look at these heat maps, which clearly show a much evolved plate discipline, it appears this change is positive for Baez. This is a positive trend, and if it continues into 2018, if he can continue to shrink his zone, then he could become a serious offensive threat.

Conclusion

The adj stat appears to sum up plate discipline in a single number, and presenting this data as a heatmap appears to show a change in approach at the plate. Together they may be useful tools going forward. For now, I cannot supply you with heatmaps for every player, but I can offer you the raw adj numbers.  I am sorry the naming convention is so horrendous, maybe I will think of a better labeling system in the future. You can view the full data set used in this article here.