26 January 2017

A Quick Glimpse At Exit Velocity

Baseball statistics have gotten increasingly complicated over time. Once upon a time, people primarily used counting stats like RBIs, home runs and stolen bases. Some of the most complicated statistics were stats like batting average, fielding percentage and ERA basically requiring little more math than average division. Then, there were statistics like FIP, wOBA and WAR. These are difficult to calculate, but are reasonably simple to understand. A higher wOBA and WAR is better than a lower one. Now, with data from Statcast and Pitch Fx available to the public, it allows us the creation of new statistics that are difficult to both calculate and understand. One of these metrics is exit velocity.

Exit velocity is a relatively new stat that I believe was introduced with the creation of Statcast. It measures how the speed of a baseball after it is hit by a batter. According to Sports Illustrated, the Rays solely use exit velocity to measure batter performance. It’s also popular with players. Jake Lamb has said that he thinks that exit velocity tells you exactly what you need to know and that he likes this stat even though he generally doesn’t care for stats. It probably didn’t hurt that this stat ranked him highly.

However, it can be difficult to understand. Take ERA for example. There’s no special bonus for giving up a low amount of runs just like there’s no special penalty for giving up a large amount of runs. Each run allowed per nine runs increases a pitchers ERA by a run. Each double hit increases a batters OPS the same as every other double. In a word, these statistics are linear. You put all the numbers into a formula and it spits out an answer that values each result similarly.

It isn’t clear that this would necessarily be the case for exit velocity. If one hits a pitch lightly, then it means that a fielder may have to charge the grounder and make a throw to first while running to get the out. Depending the game situation, the fielder’s ability or the runners speed, a fielder may decide not to make the throw. If he does make the throw, there’s the possibility that the runner will beat the throw or that it will result in an error. In contrast, a grounder hit at a normal speed will likely be easily fielded with enough to throw for first for the out. When hit into the air, it will likely result in a routine fly ball. If exit velocity isn’t linear, then it means that we’ll need to come up with different ways to understand it.

My hypothesis is that graphing exit velocity by wOBA would result in a parabolic shape. Balls with high exit velocity would have the highest average wOBA. Those with the lowest exit velocity would have a medium wOBA, and those with average velocity would have the lowest wOBA. In order to test my hypothesis, I downloaded a summary table created by Baseball Savant that has exit velocity and wOBA and graphed my results. It turns out that I was somewhat right and wrong.

The chart below shows the average wOBA by exit velocity as measured in miles per hour. The graph isn’t a parabola per se, but there’s definitely some similarity. Batters do extremely well when they hit the ball over 100 mph. From 91-99 mph, their performance begins to decrease significantly. In the sample measured by the summary table, batters had a .542 wOBA when they had an exit velocity of 99 mph and a wOBA of .257 when they had an exit velocity of 91 mph. Batter wOBA continues to drop based on exit velocity until batters hit the ball at 35 mph. Once exit velocity drops to 35 mph, wOBA increases with it to the .350 mark. It seems that hitting the ball very lightly may be better than hitting the ball 90 mph. Here’s the chart.

This next chart shows average wOBA based on exit velocity as measured in groups of either 5 or 10 mph (only for 50-59 and 60-69 mph). This chart is probably a bit clearer to understand than the first one and clearly shows how a small increase in velocity leads to a large increase in production only if the ball is hit with an exit velocity of over 95 mph. It does seem to show that batters have a small bump in performance if they hit the ball between 60-74 mph, but this could just be a quirk of the data that wouldn’t show up if we used a different sample.

These charts show why exit velocity is different than other more standard stats. There’s a huge bonus for hitting the ball really hard. But if you hit the ball at an average speed, then it doesn’t really matter much if you hit the ball 60 or 70 mph. Your results will likely be substandard. There is also evidence that suggests that an exit velocity of 20-34 mph will result in better results than hitting the ball at 60 mph. To be fair, there were only 119 balls put into play between 20-34 mph and Statcast has difficulty measuring pitches hit that weakly. This may just be the result of an error in the data and may not actually exist in real life.

This is why there’s only a .561 correlation between exit velocity and wOBA when looking at all balls put into play measured by Statcast by all batters. It’s true that a hard hit ball is better in general than a softly hit ball, but that doesn’t make it true in all instances. As such, this is one reason why average exit velocity has limited value.

Instead of looking at average exit velocity, it is therefore important to look at exit velocity in bins. Pitches hit with an exit velocity of 105+ mph should be considered highly valuable. Pitches hit with an exit velocity between 100 and 105 should also be considered successful at bats. Pitches hit between 95-99 mph are not as valuable but still likely to be somewhat productive. Anything lower than 95 mph should probably be ranked as needing improvement.

Potentially, this may mean that a batter with an average exit velocity of 85 mph can be less valuable than a batter with an average exit velocity of 75 mph. If the batter with an average exit velocity of 85 generally puts the ball into play with an exit velocity between 80 and 90 mph, then he’s not going to be productive. However, a batter with an average exit velocity of 75 mph but puts a lot of pitches into play with an average exit velocity of over 100 mph or under 40 mph could be extremely useful. Especially if he’s a quick runner and can beat out a tough throw to first. In addition, it’s possible that an acceptable exit velocity depends on the type of pitch thrown.

Some of the new statistics available to the public are excellent and can help advance our understanding of baseball but don’t necessarily work like statistics have in the past. As a result, we’ll have to use different methodologies to gauge their value and what they’re telling us about baseball players.


Unknown said...

My fear is that missing data from Statcast is skewing your results. Statcast still misses almost 50% of popups (which, as you've written before have a wOBA of .022), and roughly 15% of groundballs (which you had at .251 wOBA). It's worth noting that the most missed groundballs are those with extreme negative launch angles, i.e., straight down into the ground, which likely fit into your <35 mph bucket.

wOBA on batted ball type from here:

Statcast miss % from here:

Matt P said...

Missing data from Statcast is highly likely to be skewing my results. I mentioned that in this post, but I probably should have put it in its own paragraph to give it more prominence. I forget the exact numbers, but the Statcast wOBA was higher than the actual wOBA by a decent sized amount.

I'm not convinced that a better data set would show that batters have minimal success in the 20-35 mph bucket similar to that in other buckets of low velocity, but I agree it's a very plausible possibility. I think that Statcast would have to be missing about 80% of pitches in this bucket for that to happen which is very possible.

If that's true though, then the data still won't be linear though. The data will be linear with a very low slope when pitches are hit with an exit velocity of roughly 30 mph to roughly 94 mph and then production will increase significantly for gains in velocity once a pitch is hit over 95 mph (roughly).

Averages will still be of questionable value, while exit velocities above a certain point will be valuable. With better data, it may be possible for someone to determine whether there's a certain exit velocity where hitters suddenly become more productive and if so why that's the case.

If that is the case, then this could be a highly significant stat based on its behavior for minor league players also. In theory, you'd be able to use it to grade players in the minors/college/high school and see whether they're making good contact or not and thereby being able to control for the differences in talent.

Well, maybe. It would be quite the challenging process.

Pip said...

Last year, much was made of Joey Rickard's exit velocity, and he was maligned as being the recipient of a bunch of benevolent bloops.
Regardless of whether that is true or not, is exit velocity most easily improved by just getting stronger, or by developing a better eye, or something else?

Matt P said...

Excellent question. I don't know. With one year of data, I don't think anyone knows.

vilnius b. said...

Re: Pip's question. Doesn't the Statcast leader board for average exit velocity tell us something intuitive? Wouldn't you expect bigger and stronger players to have a higher average exit velocity?
That seems like simple physics. I'm not surprised that guys like Judge, Stanton, Nelson Cruz, Trumbo or Miggy are at the top of the leader board for 2016---I doubt that anybody would be.

I just checked out the top 15-16 hitters for that statistic and---with the exception of Keon Broxton of the Brewers---all of them are big guys. So it looks more like a way of measuring strength and mass. And even Broxton isn't a small guy; he's listed as being 6'3" & 195 lbs.

So, I guess the stat has limited utility (as Matt said). It doesn't take a genius to figure out that bigger and stronger guys should have higher exit velocities when they square up the ball.

Pip said...

That would seem to indicate that if Joey could add 20 pounds, he would have a higher exit velocity. However, a ball that hit into the ground or straight up at 100 miles
an hour would probably count as zero, so a good eye should count for something as well.
Thank you for supplying the additional information. Did it say where Jose Altuve is on that list? or Ichiro Suzuki? They are neither big nor bulky but are very successful.

vilnius b. said...

I don't see Altuve on that list, but then they only list the top 50. (Perhaps there's another way to see an expanded list but I've taken my neuropathy medication and I can't figure that out.)

However, one thing that Matt said in his article appears to be true. When the average exit velocity dips below 94 (and esp. under 93) you see a fair number of batters that weren't very good hitters in 2016 or didn't get many at bats (e.g., R. Weeks, A. Lind, J. Smoak, Bruce Maxwell and Brandon Snyder).

Roger said...

Does anybody think that averaging in missed swings as "0 exit velocity" would be a good way of assimilating strikeout percentage into the value of a hitter? Also, does this data include foul ball exit velocity?

I would bet that the "bump" in data around 60-70mph is where most "seeing eye groudballs" lie as well as the low velocity being where bunts and swinging bunts lie.

Matt P said...

"Wouldn't you expect bigger and stronger players to have a higher average exit velocity?"

His question was whether it was easier to improve exit velocity by getting stronger or by having a better eye. Both are likely to have an impact. After all, hit the ball in the right spot and it will go faster. I don't know which is more important, but would agree that both will have an impact.

"So it looks more like a way of measuring strength and mass."

If a team really thought it was doing that, then this stat has tremendous value. You could build a model that determines the percent of pitches a player should hit over a certain speed based on his size and skill level (I'd think an 20 year old in LoA that's 6'3 and 230 should be expected to do worse than a player who is 25, the same size and in the majors). Then you can tell whether a player is advanced or not based on his size.

"I would bet that the "bump" in data around 60-70mph is where most "seeing eye groundballs" lie as well as the low velocity being where bunts and swinging bunts lie."

That could be. With better data, I could probably give a better answer. Frankly, if I was working for a team, that would be a question I'd leave for the scouts to figure out.

vilnius b. said...

Matt: there's no doubt that having a better eye should have an impact on average exit velocity---for the reasons you stated. But I chose not to respond to that part of the question since it seemed self-evident to me. My bad.

As for your second rejoinder: Yes, but only for certain positions (corner OF, 1B, 3B, C). With the renewed emphasis on defense in baseball, if I were a GM I wouldn't want my shortstop, esp. those who were leaders in the DRS category the last couple of years (like Lindor, Simmons, Russell or even Crawford) to bulk up too much. Added weight would reduce range, though it might add arm strength.

The same holds true for CFs. While players like Trout bring tremendous offensive value to their teams, the best at that position as measured by DRS (I don't quite trust UZR yet) are not all that big (Kiermaier, Inciarte, Hamilton, etc.).

I'm reminded of what Earl Weaver said so many years ago: "A defensive run saved is the same as a run scored." A lot of teams are beginning to embrace that philosophy again.

Matt Perez said...

Suppose you measured 100 balls put into play against a major league pitcher by Ryan Flaherty and Lebron. Ryan is smaller than Lebron, but would probably have a higher average exit velocity on balls put into play and would put more pitches into play at over 95 mph. This is because Ryan is a baseball player and Lebron isn't.

So, you could use this stat to show, while controlling for size, which players hit the ball hard for their size and which ones didn't.

vilnius b. said...

I'd be surprised if LeBron James could even put the ball in play against a major league pitcher!

I know he's one of the greatest athletes of our time but as Michael Jordan demonstrated more than two decades ago, learning how to hit a baseball isn't something you can just pick up again. And LeBron, unlike Michael, didn't play H.S. baseball. (At least to the best of my knowledge.)

But wow! Think what he might have accomplished as a baseball player if he had taken up the game at an early age! Which brings me to one of my great disappointments late in life: why did so many African-Americans turn away from this great sport? I've heard a multitude of explanations for this unhappy trend over the last four decades, but none of them is completely satisfying to me.

But I digress, so please excuse me. Thanks for the article about exit velocity. For an old fart like me it's so hard to digest all of the new metrics.

FearItself said...

"I would bet that the "bump" in data around 60-70mph is where most "seeing eye groudballs" lie as well as the low velocity being where bunts and swinging bunts lie."

You might be right on those ground balls, and it wouldn't surprise me if a lot of Texas-leaguers are in that bump, too, but I'm almost certain bunts and swinging bunts would fall in the peaks in wOBA between roughly 23 and 34 mph.