Exit Velocity, Part III: Applying Meaning to the Data

After first demonstrating that batted ball exit velocity matters, and then establishing that it might stabilize rather quickly and represent an actual repeatable skill, the next step in our exploration of the data is its application. We want to find something that’s predictive and could possibly provide clues for future performance. In the second part of this series, we looked at a lot of relationships between first- and second-half data to determine if exit velocity is a repeatable skill. To attempt to find meaning in the data, we will again use the numbers we have for the first and second halves with a view towards identifying some meaningful information.

Looking for potentially predictive information, the simplest thing to do is look at the overall outcome — in this case, second-half production — and see if there is anything in the first half which might have portended the numbers from the second. In Part II, a scatter plot of first- versus second-half wOBA was used to show the relationship between halves. Here is that graph again.

wOBA- 1ST HALF TO 2ND HALF CORRELATION

Now here is one with first-half exit velocity and second half wOBA.

1ST HALF EXIT VELOCITY AND 2ND HALF wOBA

This relationship is not overly strong, but what it is interesting is that it is just as strong as the image that preceded it and which featured first-half wOBA verus second-half wOBA. There’s probably not enough here alone to do much with it, but for a comparison with other first-half stats and how they relate to second-half wOBA, see the chart below.

Comparing 1st Half Stats to 2nd Half wOBA
1st Half 2015 r r^2
BB% 0.43 0.18
ISO 0.36 0.13
Exit Velocity 0.34 0.11
wOBA 0.32 0.10
SLG 0.30 0.09
OBP 0.29 0.08
K% 0.08 0.01
BABIP -0.05 0
BA 0.01 0

Of the statistics above, first-half walk rate has the strongest relationship with second half wOBA, and given how quickly walk rate stabilizes and how important walks are to overall production, this cannot be too surprising. Taking a slightly different tack than above, the next chart documents the relationship between first-half exit velocity and second-half statistics.

Comparing 1st Half Exit Velocity to 2nd Half Stats
2nd Half r r^2
ISO 0.48 0.23
SLG 0.40 0.16
BB% 0.34 0.12
wOBA 0.34 0.11
OBP 0.21 0.05
K% 0.18 0.03
BABIP -0.05 0
BA 0.04 0

These results seem interesting, but as far as application goes, it is difficult to tell what to do with them. As a result, I thought I might try a different approach. There is a decent relationship between exit velocity and wOBA, so the next step here was to try and determine what happens for individual players when that relationship does not exist. To that end, I created IQ-type scores for exit velocity and wOBA from the first half of last season based on the averages of the 130 players in the sample. In each case, I assigned a figure of 100 to the sample’s average and then, for each standard deviation (SD) up or down, added or subtracted 15 points.

Once the IQ scores for both stats were calculated, I subtracted the IQ score for exit velocity from the IQ score for wOBA to find the players with the biggest disparities. The players in the chart below outperformed their exit velocity by the biggest margin in the first half of last season.

First Half Exit Velocity Overperformers
wOBA Exit Velo wOBA IQ Exit Velo IQ Diff
Bryce Harper 0.482 90.64 158 111 47
Anthony Rizzo 0.407 88.37 129 94 35
Starling Marte 0.337 85.93 102 76 27
Charlie Blackmon 0.356 87.11 110 84 25
Brian Dozier 0.357 87.44 110 87 23
Brett Gardner 0.373 88.41 116 94 22
Adrian Gonzalez 0.371 88.48 115 95 21
Buster Posey 0.377 88.79 118 97 21
Jhonny Peralta 0.355 87.91 109 90 19
Victor Martinez 0.313 85.84 93 75 18

For the most part, we have some really good offensive performances with below-average exit velocities, although Bryce Harper’s offense was just so phenomenal, he even outperformed an above-average exit velocity. At the other end, Victor Martinez did not perform well offensively, but his exit velocity was so poor he still managed to make the top ten of this list.

Now here is that same group of players with their wOBA in the first and second halves.

Exit Velocity Overperformers wOBA Difference
wOBA wOBA Diff
Bryce Harper 0.482 0.438 -0.044
Anthony Rizzo 0.407 0.356 -0.051
Starling Marte 0.337 0.337 0
Charlie Blackmon 0.356 0.331 -0.025
Brian Dozier 0.357 0.280 -0.077
Brett Gardner 0.373 0.271 -0.102
Adrian Gonzalez 0.371 0.333 -0.038
Buster Posey 0.377 0.346 -0.031
Jhonny Peralta 0.355 0.277 -0.078
Victor Martinez 0.313 0.262 -0.051
AVERAGE 0.373 0.323 -0.050

On average, this group dropped 50 points in wOBA from the first half to the second half last year. While some of the performances were just so good that they had to come down, the fact that the group went from great to average in one half says something. The only player able to buck the trend was Starling Marte and he upped his exit velocity by 3.1 mph in the second half. All the players except for Marte and Martinez, who also had a jump in exit velocity, kept their exit velocities close to their first-half numbers. Interestingly, No. 11 on the list was Joey Votto, who had a great second half and also saw a big spike in exit velocity between halves.

There was not a discernible trend as the IQ scores narrowed. Once the major outliers at the top were removed, the differences in wOBA looked fairly random until we got to the bottom of the list.

First Half Exit Velocity Underperformers
wOBA Exit Velo wOBA IQ Exit Velo IQ Diff
Michael Taylor 0.286 89.69 83 104 -21
Ryan Braun 0.356 93.69 110 133 -24
Gregory Polanco 0.285 90.21 82 108 -25
Christian Yelich 0.317 91.86 95 120 -25
Yoenis Cespedes 0.349 93.88 107 135 -28
Ian Desmond 0.258 89.42 72 102 -30
David Ortiz 0.323 92.88 97 127 -30
Robinson Cano 0.287 91.21 83 115 -32
Mark Trumbo 0.307 92.31 91 123 -32
Wilson Ramos 0.290 91.67 84 118 -34

Not unlike the list above, we see performances from players that could not help but improve in the second half, but we see above-average exit velocities throughout as well as a couple above-average offensive performances from Ryan Braun and Yoenis Cespedes. Here is how this group did in the second half.

Exit Velocity Underperformers wOBA Change
1st Half wOBA 2nd Half wOBA Diff
Michael Taylor 0.286 0.263 -0.023
Ryan Braun 0.356 0.381 0.025
Gregory Polanco 0.285 0.324 0.039
Christian Yelich 0.317 0.374 0.057
Yoenis Cespedes 0.349 0.389 0.040
Ian Desmond 0.258 0.337 0.079
David Ortiz 0.323 0.448 0.125
Robinson Cano 0.287 0.395 0.108
Mark Trumbo 0.307 0.351 0.044
Wilson Ramos 0.290 0.233 -0.057
AVERAGE 0.306 0.350 0.044

While Wilson Ramos and Michael Taylor never quite got on track, the group as a whole made a big step forward, improving by 44 points in wOBA. Unlike the group at the positive end, the underperformers kept going for a bit as far as seeing improvement in the second half. The next 30 players on the list above improved by 28 points in wOBA on average, with 24 of 30 players seeing improvement in wOBA, improving from .319 in the first half to .347 in the second half.

We could get these same results if we simply looked at the top ten and bottom ten in wOBA from the first half. Players who perform extremely well are less likely to do so while players who perform very poorly are likely to see a rise in their performance, but to be able to pluck out other players who are likely to do the same could be significant. Of the players in the overperforming list above, only two (Harper, Rizzo) were in the top 10 of the first half for wOBA and only Buster Posey was in the top 20. In the underperforming group, only Desmond was in the bottom 10 for wOBA, with just Taylor, Cano, Ramos, and Polanco among the bottom 35 of wOBA in last year’s first half.

The correlation coefficient between the difference in wOBA from the first half to the second and the difference in first-half IQ scores was -.37 with an r-squared of .14. And perhaps also of interest, the correlation coefficient for the difference in wOBA from the first half to the second and the difference in exit velocity from the first half to the second half was .49 with an r-squared of .24. It’s worth mentioning the standard caveat again: we do not have all the data necessary to form firm conclusions. The data should get better, and adding another year should add clarity. Whether that makes this data look more sturdy or more random we do not yet know. For one year, however, the massive overperformers relative to exit velocity took a dive in the second half, and almost all underperformers saw their second-half performances go up. I am anxious to see if this trend continues next year.

For those concerned, there are no current plans for a Part IV.





Craig Edwards can be found on twitter @craigjedwards.

6 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
bwmember
6 years ago

I really appreciate this series of posts for some validation to the strategy I only guessed was useful last year. Unfortunately, trusting velocity was one area that I felt my league hadn’t quite come around on, and now you’re going to ruin that 🙂