Two Short Studies: Groundballs Pitchers & StatCast Projections

Groundball Pitchers Suppressing ERA

I am going to play with fire and refute a Dave Cameron comment. In a recent article about Brad Ziegler, Cameron said:

So Ziegler basically breaks every mold you can think of. And he even breaks our models. His career FIP is 3.38, but his career ERA is 2.44, almost a full run lower. Part of that is that groundball pitchers get to count more of their runs as unearned because there are more errors on groundballs than on flyballs, so ERA systematically is biased in favor of groundball pitchers.

After the work I just did on pERA, I was worried about its validity. Previously, I found that groundball and flyball pitchers exponentially suppress their ERA as they move to extreme ends of the batted ball spectrum.

I needed to determine if the extreme ERA and FIP difference at high groundball levels was exacerbated by fielding errors.

I studied starting pitchers from 2002 to 2016 (minimum of 100 IP). I took their RA (all runs allowed per nine innings pitched) minus ERA and compared the difference against their groundball rate. I got a positive correlation and here are the average results using the calculated equation.

Groundball Rate vs RA-ERA
GB% Projected RA-ERA
25% 0.28
30% 0.30
35% 0.32
40% 0.34
45% 0.36
50% 0.38
55% 0.40
60% 0.42
65% 0.45
70% 0.47

The r-squared between the two values was .02 so the correlation is not strong. On average, the difference between RA and ERA increased .004 for every 1% point increase in groundball rate. Not a huge increase, but measurable.

Using the same dataset, I decided to bucket the data into 5% point groups. Then, I found the median value for each of these groups and here are the results.

Groundball Rate vs RA-ERA
GB% Median RA-ERA
<30% 0.187
30% to 35% 0.334
35% to 40% 0.289
40% to 45% 0.321
45% to 50% 0.338
50% to 55% 0.367
55% to 60% 0.396
> 60% 0.377

The outcomes from both analysis are similar. Some correlation, but not strong.

Some ERA suppression is from unearned runs caused by fielding errors but the difference from an extreme flyball to groundball pitchers is 0.40 ERA and only 0.20 difference from an average pitcher. Unearned runs don’t come close to explaining the historic 1.00 point difference seen by pitchers with 60% ground ball rates. Sorry Dave.

Using StatCast for xISO and xHR/FB%

MLB.com’s StatCast data makes exit velocity values publicly available. With these values, some “x” values (expected) have been created to better understand hitters. The problem with StatCast data, one out every eight batted balls is missing. Most of the missing data consists of weak infield groundballs and popups. I corrected for the missing data and used the new values to create a couple of expected metric.

I used Statcast’s average exit velocity (EV) and batted ball distance to come up with an expected isolated power (xISO) and home per flyball rate (xHR/FB%). To get these, I ran a linear correlation between the values to get a prediction equation. Here are the r-squared values for people who are into those types of things.

Correlation: r-squared
EV vs ISO: .27
EV vs HR/FB%: .29
Batted Ball Distance vs ISO: .44
Batted Ball Distance vs HR/FB%: .23

I am not surprised to see non-ideal results with player sample sizes as low as 30. With the equations, I was able to generate these reference tables.

EV and xISO and xHR/FB%
Average EV HR/FB% ISO
75 -2.6% 0.032
76 -1.1% 0.044
77 0.3% 0.056
78 1.7% 0.068
79 3.1% 0.080
80 4.5% 0.092
81 5.9% 0.104
82 7.4% 0.116
83 8.8% 0.128
84 10.2% 0.140
85 11.6% 0.152
86 13.0% 0.164
87 14.4% 0.176
88 15.8% 0.188
89 17.3% 0.201
90 18.7% 0.213
91 20.1% 0.225
92 21.5% 0.237
93 22.9% 0.249
94 24.3% 0.261
95 25.7% 0.273
96 27.2% 0.285
97 28.6% 0.297
98 30.0% 0.309
99 31.4% 0.321
100 32.8% 0.333

 

Average Batted Ball Distance and xISO and xHR/FB%
Average Batted Ball Distance xHR/FB% xISO
140 -0.8% 0.001
145 0.2% 0.013
150 1.2% 0.026
155 2.2% 0.038
160 3.2% 0.050
165 4.3% 0.062
170 5.3% 0.075
175 6.3% 0.087
180 7.3% 0.099
185 8.3% 0.112
190 9.3% 0.124
195 10.3% 0.136
200 11.3% 0.149
205 12.3% 0.161
210 13.3% 0.173
215 14.4% 0.185
220 15.4% 0.198
225 16.4% 0.210
230 17.4% 0.222
235 18.4% 0.235
240 19.4% 0.247
245 20.4% 0.259
250 21.4% 0.272

Additionally, I added the expected values to the online spreadsheet.

Finally, I found the range of possible values by first averaging the two xHR/FB and xISO values. Then I subtracted the averaged expected values from the actual values. The difference’s standard deviations worked out to be 0.046 for ISO and 5.7% for HR/FB. Again not great results, but a starting point for now. I plan on streamlining the process for improved results as more factors, like flyball rate, needs to be included.





Jeff, one of the authors of the fantasy baseball guide,The Process, writes for RotoGraphs, The Hardball Times, Rotowire, Baseball America, and BaseballHQ. He has been nominated for two SABR Analytics Research Award for Contemporary Analysis and won it in 2013 in tandem with Bill Petti. He has won four FSWA Awards including on for his Mining the News series. He's won Tout Wars three times, LABR twice, and got his first NFBC Main Event win in 2021. Follow him on Twitter @jeffwzimmerman.

6 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
scotman144member
7 years ago

GB pitchers suppressing ERA: double plays surely help explain the effect no? They’ll allow more baserunners but will allow lower SLG/ball in play (less runner advancement) and will generate many more DP opps.