Which Starters Would Be Good Relievers?

I’ve been thinking about Wade Davis and Joe Kelly lately. Davis started 29 games in back to back seasons for the Rays in 2010 and 2011, striking out fewer than six batters per nine and putting up FIPs and xFIPs in the neighborhood of 4.70. The next season, the Rays moved Davis to the bullpen where he struck out more than 11 batters per nine and produced a 2.78 FIP. After trading for Davis that offseason, the Royals tried Davis at starter again for much of the 2013 season with poor results, and they’ve put him the bullpen ever since, where he’s again become dominant.

Davis and Kelly aren’t the same pitcher. Kelly throws much harder than Davis did as a starter and has a four-pitch repertoire that includes a change-up. However, both pitchers use a four-seamer, a slider or cutter, and a curveball and have most of their success with their fastballs. Meanwhile, Kelly shows similar and similarly underwhelming strikeout rates as Davis did as a starter. Kelly has had FIPs and xFIPs over 4.00 in each of his four seasons. That’s 461.2 innings of mediocre results, but the Red Sox seem to be planning to rely on Kelly as a starter again in 2016.

Without any real justification, I’ve been convinced for some time that Kelly belongs in the bullpen and could be an exceptional reliever. Finally, I’ve decided to see if I could find some statistical proof that that is the case.

Obviously, some starters succeed in a transition to the bullpen because they see an increase in velocity. That is not something I believed I could trivially predict, so I instead decided to look at the split between a starter’s performance his first time through the batting order and his subsequent times through the order. My thought was that many starters have success the first time a batter sees them on a given day, but if their performance drops off dramatically when they face the same batter two and three times, they might be best served in the bullpen where they would never have to face the same hitter twice in one day.

To test this, I compiled starter xFIPs in their first times through the order and in all subsequent times through the order. I relied on three-year samples with at least 20 total starts, and I excluded pitchers with xFIPs under 4.00 in their subsequent times through the order under the theory that they were good enough to start even if they were experiencing a decline. So the 2011 leaderboard, as an example, would show Davis’s xFIP splits from his 64 combined starts between 2009 and 2011. Here are the starters with the biggest decline in performance after their first time through the order through the 2011 season:

Reliever Candidates Among Starters, 2011
Pitcher GS xFIP First Time xFIP Subs. Times Diff
Jon Niese 61 2.85 4.24 1.39
Scott Baker 83 3.38 4.52 1.14
Rick Porcello 89 3.55 4.69 1.14
Jose Contreras 23 3.33 4.43 1.10
Jaime Garcia 60 2.95 4.04 1.09
Jordan Zimmermann 49 3.24 4.28 1.04
Michael Pineda 28 3.24 4.19 0.95
Jered Weaver 100 3.31 4.24 0.93
Alexi Ogando 29 3.27 4.17 0.90
Travis Wood 35 3.84 4.68 0.84

As it turns out, Davis does not show up. In fact, he improved slightly in his subsequent trips through the batting order from a 4.58 xFIP to a 4.48 xFIP. However, there were a few names that stood out to me. Alexi Ogando transitioned from starter to reliever in 2012 and saw his strikeout rate increase from 6.7 to 9.0 per nine. Jose Contreras enjoyed 3.27, 3.10, and 2.73 FIPs as a reliever between 2010 and 2012 after back-to-back seasons with FIPs over 4.00 as a starter. Travis Wood only just became a reliever in 2015, but when he did, his strikeout rate skyrocketed from 7.6 to 10.6 per nine.

There is a fair bit of overlap in back-to-back years because of the three-year trail I’m using, so let me jump ahead to 2013.

Reliever Candidates Among Starters, 2013
Pitcher GS xFIP First Time xFIP Subs. Times Diff
Scott Baker 24 3.26 4.76 1.50
Joe Kelly 31 3.23 4.20 0.97
Michael Pineda 28 3.24 4.19 0.95
Lucas Harrell 56 3.47 4.37 0.90
Dallas Keuchel 38 3.63 4.50 0.87
Matt Moore 59 3.20 4.04 0.84
Jeremy Hefner 36 3.46 4.28 0.82
Scott Diamond 58 3.70 4.48 0.78
Chris Narveson 30 3.50 4.24 0.74
John Lackey 57 3.52 4.25 0.73

Joe Kelly. Meanwhile, no one else on this list has had a significant opportunity to relieve, but Brian Duensing just missed at No. 15, and he increased his strikeout rate from around 6.0 to 8.3 per nine when he became an exclusive reliever in 2013.

Four good examples hardly make a definitive case, and even if you stretch this back to earlier seasons—Matt Belisle was fifth in 2009, Matt Albers was 16th in 2007, and Dustin Hermanson, Scott Downs, and Derek Lowe were second, third, and 11th in 2005—there may be evidence of this simply because it’s easier to be a reliever than a starter and most pitchers improve with the transition.

Still, if you are interested by this as I am, here is a list of starters with at least 15 starts between 2013 and 2015 whose xFIP the first time through the order is at least 50 points better than it is through subsequent times through the order and whose subsequent xFIP is over 4.00:

Reliever Candidates Among Starters, 2015
Pitcher GS xFIP First Time xFIP Subs. Times Diff
Yusmeiro Petit 20 2.39 4.17 1.78
Tony Cingrani 30 2.67 4.28 1.61
Tsuyoshi Wada 20 2.86 4.23 1.37
Zack Wheeler 49 2.93 4.19 1.26
Trevor May 25 3.09 4.33 1.24
Chris Heston 32 2.93 4.12 1.19
Jake Odorizzi 63 3.22 4.20 0.98
Derek Holland 48 3.23 4.21 0.98
Lucas Harrell 25 3.81 4.78 0.97
Chad Bettis 28 3.34 4.30 0.96
Anthony DeSclafani 36 3.23 4.18 0.95
Jeremy Hefner 23 3.28 4.19 0.91
Ryan Dempster 29 3.34 4.21 0.87
Vance Worley 35 3.28 4.12 0.84
Scott Diamond 24 3.88 4.70 0.82
Carlos Villanueva 20 3.61 4.42 0.81
Juan Nicasio 46 3.67 4.48 0.81
Brandon Morrow 21 3.36 4.15 0.79
Jon Niese 83 3.25 4.03 0.78
Ivan Nova 41 3.47 4.23 0.76
Drew Pomeranz 23 3.45 4.21 0.76
Henderson Alvarez 51 3.30 4.01 0.71
Jonathan Pettibone 20 3.73 4.43 0.70
Bronson Arroyo 46 3.32 4.00 0.68
Brett Oberholtzer 42 3.70 4.37 0.67
Shane Greene 30 3.34 4.01 0.67
Chris Rusin 35 3.84 4.47 0.63
A.J. Griffin 32 3.66 4.24 0.58
Andy Pettitte 30 3.48 4.05 0.57
Jason Vargas 63 3.73 4.29 0.56
Trevor Cahill 45 3.48 4.01 0.53
Yohan Flande 20 3.68 4.21 0.53
Zach McAllister 40 3.61 4.14 0.53
Robbie Ray 29 3.64 4.17 0.53
Kyle Gibson 73 3.54 4.05 0.51
Eduardo Rodriguez 21 3.63 4.14 0.51
Brad Hand 30 3.83 4.34 0.51

There are a lot of intriguing names on this list. Yusmeiro Petit and Tony Cingrani have already transitioned to the bullpen, and Trevor May appears to be headed that way. Chris Heston could land in the pen when Matt Cain is ready. Juan Nicasio has wowed this spring for the Pirates, but he could easily land in the bullpen instead of the rotation. I like him as a sleeper. Drew Pomeranz could be in the mix for saves behind Fernando Rodney in San Diego. Finally, Jon Niese is on the list basically every year. I had never spent much time thinking about Niese before, but now I’m desperate for the Pirates to promote Tyler Glasnow, Jameson Taillon, and whoever else it takes to bump Niese to the pen so I can see how he does. Come on Neal Huntington, make it happen!

Scott Spratt is a fantasy sports writer for FanGraphs and Pro Football Focus. He is a Sloan Sports Conference Research Paper Competition and FSWA award winner. Feel free to ask him questions on Twitter – @Scott_Spratt

newest oldest most voted
some guy
some guy

Really cool article.

Where did you pull this data? It doesn’t appear to be available on Brooks Baseball. Unless I missed something BB doesn’t put xFIP and times through order in the same chart…