Can Statcast Help Identify Future Relief Pitcher Success?

Last week I posted the year to year correlations for xStats and their standard variants, and it came up with a few interesting results.  The xStats variants were much more consistent year to year, for better or worse, and in general they were better at predicting future performance. Not by much in some cases, but hey, every bit helps, right?  It made me curious how it may translate to groups of players with smaller sample sizes, so this week I’ve taken these stats to relief pitchers, with those year on year correlations in mind.  Yes, it is frustrating that we only have two seasons to look at, but this is the best we have at the moment so let’s see where it gets us.

As you might remember, vertical launch angle was very consistent (.75) between 2015 and 2016 for all pitchers, and as it turns out this holds true for every innings limit you can imagine.  Whether you want to talk about guys with 30 innings, 200 innings, or anything in between.  Vertical angle appears to stabilize fairly quickly.  So, that begs the question, how does vertical launch angle change batter performance?  Hopefully this chart will answer your questions.

hr-slg-avg-vlaunch

Between roughly 10 degrees and about 35 degrees batted balls have high value, with batting average peaking around 13 degrees, slugging around 25 degrees, and home runs around 27 degrees.  So, if we know vertical launch angle is stable between seasons, and batted balls between 10 and 35 degrees are bad (for the pitcher), then perhaps aiming for pitchers who have average launch angles outside of that zone would be ideal.

Next, K/9 and K% are both solid between seasons (.74 and .72).  This should be no surprise, we’ve long known that strikeout rates stabilize quickly and hold steady over long periods of time.  I’ll limit to pitchers who have above average strikeout rates.

Finally, xAVG is strong between seasons, especially for these smaller sample size pitchers.  It appears that singles and home runs are easier to predict with smaller samples, and as the samples grow doubles become a bit easier.  Triples are always difficult to predict, which makes a bit of sense as they are often the result of defensive misplays.  xAVG is good between 2015 and 2016 for relief pitchers (.53), and also did a good job predicting future batting average (.42), and as such, perhaps it would be a good idea to keep pitchers with above average xAVG in 2016 in mind.

Home runs and home run predictors seem to fall apart when pitchers have as few innings as your standard relief pitcher (30-80), which probably explains why these pitchers have such chaotic ERA and ERA estimators across seasons.  Perhaps home runs are the result of pitcher error, and it takes a larger number of pitches to fully express the frequency of these errors?  Either way, I have filtered the 2016 pitchers with the following criteria:  

Between 30 and 80 IP, average vertical launch angle not between 10 and 35 degrees, above average K% and xAVG.

 

Relief Pitchers with superior xAVG, K%, and Vertical Angle
Name IP scFIP FIP xAVG xOBP xSLG xOBA xHR K% avg EV vertical
Aroldis Chapman L 58 1.78 1.48 .170 .236 .233 .209 2 40.5% 88.3 6.9
Andrew Miller L 74.1 2.13 1.74 .165 .199 .279 .203 7 44.7% 87.4 6.7
Dellin Betances R 73 2.24 1.80 .182 .257 .285 .239 5.3 42.3% 86.7 5.5
Alexander Reyes R 46 2.41 2.73 .169 .266 .237 .232 1.5 27.5% 85.5 9.2
Matt Bush R 61.2 2.52 2.80 .227 .273 .330 .265 4 25.1% 86.4 9.7
Xavier Cedeno L 41.1 2.77 2.70 .243 .297 .338 .282 2 24.9% 88.3 9.5
Joe Biagini R 67.2 2.89 2.79 .235 .285 .329 .272 3.5 21.4% 87.7 5.6
Jeurys Familia R 77.2 2.90 2.45 .213 .290 .286 .262 3.1 26.2% 86.3 -2.6
Mark Melancon R 71.1 2.90 2.48 .233 .270 .326 .262 4.1 24.1% 85.2 5.6
Dan Otero R 70.2 2.96 2.39 .243 .274 .339 .271 3.7 21.2% 88 2.9
Kyle Barraclough R 72.2 3.07 2.17 .178 .297 .253 .253 3 36.9% 87.4 5.4
Zach Britton L 67 3.07 2.00 .197 .254 .274 .236 3 29.1% 91.6 -1.4
Will Harris R 64 3.10 2.36 .233 .278 .334 .270 3.9 27.2% 90.2 4.3
Alex Colome R 56.2 3.14 2.92 .215 .271 .333 .267 5.4 31.6% 90.5 7.1
Luke Gregerson R 57.2 3.15 3.05 .195 .263 .315 .256 5.2 29.1% 88.3 2.5
Ian Krol L 51 3.25 2.85 .228 .278 .347 .276 4.8 26.1% 88.8 4.6
Robert Gsellman R 44.2 3.25 2.62 .235 .295 .340 .280 2.7 22.8% 90.6 4
Michael Lorenzen R 50 3.34 3.37 .246 .297 .331 .281 2.5 24.5% 87.4 1.9
Carl Edwards R 36 3.35 2.93 .211 .288 .325 .269 3.2 37.7% 86.1 9.6
Nate Jones R 70.2 3.37 2.99 .230 .280 .381 .287 7.5 29.2% 91 8.9
Cam Bedrosian R 40.1 3.39 2.12 .219 .290 .333 .276 2.7 31.7% 91.8 5.9
Ken Giles R 65.2 3.39 2.83 .225 .292 .378 .292 8 35.9% 89 9.9
Justin Wilson L 58.2 3.45 3.24 .223 .280 .350 .278 5.8 26.3% 88.5 8.6
Santiago Casilla R 58 3.51 3.90 .211 .282 .341 .276 6.4 27.2% 88.1 9.3
Zach Duke L 61 3.54 2.88 .218 .307 .303 .276 3.2 26.7% 85.1 5.6
Oliver Perez L 40 3.59 3.86 .230 .322 .339 .298 3 26.0% 87 9.5
Boone Logan L 46.1 3.65 3.16 .203 .288 .346 .279 4.8 30.8% 87 8.3
David Robertson R 62.1 3.66 3.66 .209 .302 .338 .284 5.9 27.9% 87.4 9.2
Arodys Vizcaino R 38.2 3.72 3.57 .225 .332 .338 .301 3.2 27.8% 89.7 9
Marc Rzepczynski L 47.2 3.72 3.37 .227 .330 .323 .298 2.3 21.8% 87.4 1.1
Jim Johnson R 64.2 3.72 2.73 .246 .308 .361 .297 5.4 25.7% 89.7 5.3
Joseph Musgrove R 62 3.79 4.09 .237 .286 .395 .299 7.2 21.7% 89.6 7.4
A. J. Ramos R 64 3.79 2.86 .242 .338 .362 .312 4.5 26.5% 87.8 9.1
Alex Wood L 60.1 3.79 3.24 .246 .314 .388 .309 6.7 25.9% 89.4 7.1
Daniel Coulombe L 47.2 3.81 3.65 .213 .278 .356 .276 5.6 28.0% 89.2 1.7
Hunter Strickland R 61 3.81 3.17 .244 .302 .392 .305 7 22.9% 90 8.6
Cory Gearrin R 48.1 3.87 3.41 .240 .299 .373 .298 5.1 22.7% 89 6.5
Felipe Rivero L 77 3.91 3.44 .223 .308 .353 .295 8.2 28.3% 86.9 7.6
A. J. Schugel R 52 3.92 3.25 .250 .303 .396 .309 5.5 22.6% 90.1 9.9
Fernando Rodney R 65.1 3.94 3.73 .228 .333 .331 .297 4.2 26.4% 85.8 4.4
Danny Farquhar R 35.1 3.96 4.91 .232 .310 .395 .308 4.7 29.7% 87.9 7.4
Pedro Strop R 47.1 3.96 2.91 .235 .306 .361 .295 4.6 32.3% 89.6 5.6
Blake Wood R 76.2 4.00 4.15 .236 .324 .364 .307 6.6 24.6% 90.9 5.7
Cody Allen R 68 4.02 3.37 .215 .293 .377 .292 8.8 33.0% 92 9.6
Blake Treinen R 67 4.03 3.68 .222 .310 .330 .288 4.9 24.0% 87.4 1.5
Bryan Shaw R 66.2 4.07 3.96 .245 .321 .383 .312 7.3 25.2% 86.6 5.1
Jake Diekman L 53 4.16 3.55 .236 .329 .374 .313 4.9 26.8% 88.6 9.4
Dustin McGowan R 67 4.21 4.25 .212 .307 .345 .292 7.5 22.6% 87.5 3.5
Francisco Rodriguez R 58.1 4.37 3.84 .255 .319 .426 .325 7.3 22.2% 91.5 6
Rubby De La Rosa R 50.2 4.48 4.55 .244 .324 .408 .322 7 24.3% 89.8 7.8
Trevor Cahill R 65.2 4.49 4.27 .237 .335 .373 .318 6.5 23.5% 89.4 6.5
Tyler Skaggs L 49.2 4.52 3.89 .249 .326 .407 .321 5.8 23.0% 89.1 6.5
Keone Kela R 34 4.64 4.44 .246 .334 .438 .335 5 30.4% 91 6.6
Hunter Cervenka L 43.1 4.88 4.11 .212 .332 .350 .307 4.6 23.2% 86.9 9
Chasen Shreve L 33 4.91 5.63 .252 .327 .462 .344 5.9 23.6% 88.3 8.2
Anthony Swarzak R 31 5.22 6.08 .243 .285 .492 .333 8.1 25.2% 89.2 3.8
SOURCE: xstats.org
xAVG <= .255,
K% >= 21.2%,
30-80 IP

Many of the pitchers on this list are names you’d expect to see.  Andrew Miller, Zach Britton, Aroldis Chapman, Dellin Betances, Jeurys Familia, and Mark Melancon to name a few.  You may notice Kenley Jansen is missing, his average launch angle is around 19 degrees, directly in the danger zone you can see in the chart above. Of course he has managed success in that area to this point, so perhaps this sort of filtering isn’t perfect.

A few names on this list might be illuminating.  Joe Biagini, for example, while hyped a good bit by many perhaps hasn’t been given the respect he deserves at this point in his career. The rookie Robert Gsellman (pronounced like the animal “gazelle”, except with “man” at the end) is slotted into the relief role for the Mets going into 2017.  Given his prior major league success (as a starter), these stats, and the relative weakness of the Mets bullpen, perhaps he could rocket his way into a significant bullpen role.  For most fantasy leagues this isn’t interesting, but if you’re playing in a league that favors bullpen arms, keep your eye on him.

I personally love Treinen, as he has limited batters to consistently poor contact (xAVG .224 and .222, vertical angle of 1.8 and 1.5, Exit Velocity 87.9 and 87.4 mph) in the past two seasons. His strikeout rate isn’t spectacular, and perhaps never will be, but weak contact is nothing to sneeze at, especially with a good defense behind him.  Daniel Murphy, Ryan Zimmerman, and Jayson Werth may not be ideal defensive specialists, but the rest of the team should be solid enough to help carry Treinen toward success.

Are you surprised to see any names on this list?  Or missing from this list?

We hoped you liked reading Can Statcast Help Identify Future Relief Pitcher Success? by Andrew Perpetua!

Please support FanGraphs by becoming a member. We publish thousands of articles a year, host multiple podcasts, and have an ever growing database of baseball stats.

FanGraphs does not have a paywall. With your membership, we can continue to offer the content you've come to rely on and add to our unique baseball coverage.

Support FanGraphs




Andrew Perpetua is the creator of CitiFieldHR.com and xStats.org, and plays around with Statcast data for fun. Follow him on Twitter @AndrewPerpetua.

newest oldest most voted
Jim
Member
Member
Jim

Andrew, you are perpetually (sorry, couldn’t resist) doing excellent work. How can we learn more about you?