clock menu more-arrow no yes

Filed under:

Effective exit velocity: background, data and charts

New, 11 comments

Exit velocity is not created equal

Toronto Blue Jays Vs Boston Red Sox Rick Madonik/Toronto Star via Getty Images

When Statcast data first became available in 2015, the immediate emphasis was on the exit velocity that hitters generated off the bat. Intuitively, even the most casual observer of baseball understands that the harder one hits the ball, the better. It’s a simple, straightforward narrative, and as the data emerged and it was studied, the relationship was unambiguous:

EEV exit velocity bins

This chart and all others below are based on the Statcast data for the entire 2020 season, over 43,000 balls in play. Above, they’re binned by increasing exit velocity velocity into roughly groups of 1,200 balls with the average results.

Below 90 MPH, batters average consistently hit about .200 with almost no power; overall it works to abysmal production of about 27 wRC+. Production inches up from there, but it really takes once exit velocity gets to 95 MPH (incidentally, the bin where production reaches roughly average).

Batting average increases consistently, with power output absolutely exploding as exit velocity increases from that point. The magnitudes of the rise of each component is a little different, but they move in the same direction at the same inflection points and continuously to create a clean and clear story: the harder the ball is hit, the better the outcome. There may be some diminishing returns at the highest exit velocity, but the results don’t actually reverse.


Yet at the same time, it was and is equally intuitively clear that exit velocity wasn’t the only factor; that all exit velocity isn’t created equal. The first Statcast debate I remember having here was in late 2016 about Aaron Sanchez and the idea that he was due to regression because he had one of the higher rates of giving up balls at 100+ MPH.

Crucially, he also had one of the highest ground ball rates in baseball around 55%, and a lot of those hard balls are on the ground. Sure, a few more of them will go through for hits, but at worst it’s a double. No matter how hard hit, a ground ball can’t go over the fence (and often can be turned for two). It’s in the air that exit velocity really causes damage.

I’ve performed the same exercise on the 2020 data as above, except segmenting instead by launch angle:

EEV launch angle bins

Below -10 degrees, batting average tops out at .200, but from that point to the low-teens degrees it ascends rapidly to a maximum around .800. This point is what I call “pure” line drives, balls that come off the bat like bullets. Power on the other hand remains negligible through about 10 degrees, before taking off dramatically.

Unlike the exit velocity chart, not only do the two components not align, but the critical distinction is both peak and then decline. Higher is better, but only to a point, and eventually the two extremes are equally as useless. Batting average remains strong through the the mid-20 launch angles, before really tailing off. Power peaks in the high-20s, before rapidly tailing off to almost nothing by 40 degrees.

EEV launch angle wRC+

It’s worth looking at the combined effect here with wRC+ (which appropriately weights not making outs more highly than power). The dual peak is caused by the dip on what I call “intermediate liners”, well hit balls in the 17-20 range but that hang up enough for outfielders without the carry to go over their heads or the wall very often.

The bottom line though is along the zone from about 10 degrees to 30 degrees, batters do essentially equally as well production-wise, despite it looking very different at one end than the other in terms of the breakdown between power and on-base. There’s some decent production adjacent to that, but it tails off and below 0 or above 40 are essentially useless.

Clearly, any model needs to consider both exit velocity and launch angle, but the nature of the latter makes the analysis much more complicated than simple linear regression. Hence, the xwOBA model at Baseball Savant that uses a “nearest neighbours” approach to generate an expected value for each combination. This is the right approach, but the drawback to me is that the results and tradeoffs are not intuitive. For example, is it better to trade one MPH for one degree of LA?


A final point is that exit velocity and launch angle are not independent of each other, in fact there’s significant correlation (at least to a point):

EEV LA EV correlation

This should also be pretty intuitive: hitting a ball really hard almost necessarily involves squaring it up pretty well and mishitting a ball makes it hard to hit with any authority (though it is more common to square a ball up softly, off the end of the bat). Hence as launch angle increases, average exit velocity increases quickly before flattening out. Sorting by exit velocity and looking at the average launch angles produces the same trend.

But the extremes or edge cases are quite instructive, as became clear to me over time as the Statcast output became more ubiquitous and available in real time watching games. In particular, I’d see utterly routine ground balls that would register 100+ exit velocity that for a time had me convinced system had significant measurement errors because the was no way such banal contact could actually be that hard.

This is where I’d like to insert a reel of selected 2021 plays in one handy clip, but unfortunately MLB in their infinite wisdom disabled their fantastic Film Room product at the beginning of lockout this month (beyond accessing previous reels) so I can’t put one together to demonstrate. The best I can do is insert a couple below and link to a more (Kirk, Bichette, Gurriel, Dickerson, Grichuk, Vladdy):

That ball had a measured exit velocity of 107.5 MPH, and while there’s the slightest sharpness, it’s chopped so far down (-16 degrees) it ends up a banal groundout. This one by Vladimir Guerrero is even more extreme, in that it’s hit 105+ MPH at -35 degrees downward:

Conversely, consider the following balls:

Both are putatively hit softly (74 and 80 MPH respectively), but both are squared up at near optimal angles and appear quite sharply struck. Whatever their actual exit velocity, their apparent or perceived exit velocity is much higher and it’s much more effective contact than the much harder hit balls shown above.

Thinking about plays like these and crunching the data has led me to the conclusion that while exit velocity is certainly an important factor, it’s launch angle that’s more important or paramount. Exit velocity only really matters within bands of launch angles where it can be leveraged into significant outcomes. If we want to make leaderboards and rankings of exit velocity, they should be adjusted for the usefulness of that exit velocity. That is, they should be of effective exit velocity. In the near future, I will expand on this concept.