A look at luck

by The Dark on Jul 14, 2010 9:07 AM EDT

Given some of the recent discussions on how stats are related to luck, I wanted to look at how the luck of various pitchers compares, using the difference between ERA and FIP as a crude measure of pitchers' luck. I also wanted to look at the difference between starters, middle relievers, and closers. Just to establish some sort of basis, I took the top ten players in Quality Starts, Holds, and Saves, as well as anybody else with the same number of those as the tenth guy. I then calculated "Luck" as their ERA-FIP, divided by 9 to get Per-Inning Luck, multiplied by IP to get Total Luck, and then summed everyone's Total Luck, divided by Total IP, and multiplied by 9 to get a Weighted Average Luck for each group (snazzy acronym suggestions will be accepted). One caveat: by taking the top pitchers of each group, presumably I'm also getting some of the luckiest pitchers, so their WAL will presumably be better than average for their position.

The starters:
Felix Hernandez
Josh Johnson
Adam Wainwright
Roy Halladay
Tim Hudson
Ubaldo Jimenez
Roy Oswalt
Fausto Carmona
Jeff Niemann
David Price
CC Sabathia
Jered Weaver

Combined, the starters have pitched 1,513 innings. They range from the slightly unlucky Weaver (-0.15 ERA-FIP) to the very lucky Hudson (1.93 ERA-FIP). As a corps, the starters have a WAL of 0.75, suggesting that these guys are moderately lucky.

The middle relievers:
Mike Adams
Daniel Bard
Luke Gregerson
Joba Chamberlain
Kevin Jepsen
Tyler Clippard
Joe Beimel
Scott Downs
Brandon Lyon

Combined, these player have pitched 350.2 innings. Joba has been hideously unlucky, with a -3.07 ERA-WIP, while the luckiest is Daniel Bard's 1.25. On a whole, the middle relievers have a WAL of -0.27, so they are unlucky overall. However, removing just Joba from the equation would shift the WAL to 0.07.

The closers:
Joakim Soria
Heath Bell
Francisco Cordero
Neftali Feliz
Rafael Soriano
Brian Wilson
Matt Capps
Francisco Rodriguez
Matt Lindstrom
Kevin Gregg
Billy Wagner
Mariano Rivera
Leo Nunez
Jon Rauch
Jonathan Papelbon

These players have pitched a total of 554.2 innings. Nunez and Neftali are the only unlucky pitchers, with Nunez being at -0.76. Kevin Gregg is perfectly balanced, and the luckiest closer is Rivera, at 1.22. As a whole, relievers are somewhat lucky, at 0.44 WAL.

Given the limits of this look, it appears that the current top starters are luckier as a whole than the top closers, who are luckier than the top middle relievers. Given time over the next few weeks, I'll try to look at some of the worst in each category to see if the pattern is consistent that starters>closers>middle relievers for luck.

Tweet Comment 9 comments | 0 recs |

Do you like this story?

Comments

Doing it by top QS, H, SV, or whatever is biased because those players by definition should be more lucky than the rest. The average ERA-SIERA of the top 10 starters by ERA is -1.2 but by top 10 SIERA is only -0.2. Look at top 10 by FIP, and see if you get the same result. If so, it might be an issue with high K correlating with low BABIP, which is not something FIP corrects for, but which SIERA does.

by Matt Swartz on Jul 14, 2010 9:54 AM EDT reply actions

The top 10 list by FIP:
Francisco Liriano
Josh Johnson
Cliff Lee
Roy Halladay
Yovani Gallardo
Jon Lester
Tim Lincecum
Jered Weaver
Adam Wainwright
Clayton Kershaw

1,208 innings. Least lucky is Liriano, at -1.68. Luckiest is Wainwright, at +0.95. Overall WAL is +0.13, so there’s still some suggestion of luck there, although far less than the +0.75 of the starters I originally had.

That was also part of why I wanted to look at the “unlucky” pitchers (those with lots of non-quality starts, appearances without holds, or blown saves), to see if the pattern held that starters have a larger ERA-FIP than closers, who are higher than middle relievers. I did recognize that I was probably going to have most of the luckiest pitchers, so I expected all to be positive; the middle relievers being negative surprised me until I saw how unlucky Joba has been.

Honor is no substitute for victory.

by The Dark on Jul 14, 2010 11:36 AM EDT up reply actions

You may also want to look at xFIP in addition to FIP. xFIP accounts for variations in HR/FB, as generally the pitcher’s only effect on HR is their GB/FB ratio. FIP doesn’t account for pitchers who may have given up more/less HR than expected given their GB rate.

http://www.hardballtimes.com/main/fantasy/article/for-those-who-still-dont-believe-fip-is-poor-for-fantasy-analysis/

by Rujasu on Jul 14, 2010 4:20 PM EDT up reply actions

I considered that, but I picked FIP because both FIP and ERA are affected by the HR/FB ratio; by subtracting them from each other, I think that should reduce the effect of HR/FB ratio in luck, since it appears on both sides. It may not be a perfect offset, but the effects of HR/FB should be reduced by using FIP instead of xFIP. If ERA weren’t also affected by the HR/FB rate, I would have used xFIP.

Honor is no substitute for victory.

by The Dark on Jul 14, 2010 4:29 PM EDT up reply actions

I think your reasoning is flawed here. xFIP assumes a league average HR/FB, so using that would include the difference between the pitcher’s HR/FB and the average, as that’s part of luck. Using FIP, both the ERA and the FIP have the same HR/FB, so it doesn’t include the effect of HR/FB as part of luck — like you said, it would appear on both sides, so the difference is assumed to be zero.

To be clear, I’m viewing HR/FB as being part of luck here, and thus something that should be included in the WAL number.

by Rujasu on Jul 15, 2010 8:31 AM EDT reply actions

This is meant to be a reply to the post above.

by Rujasu on Jul 15, 2010 8:33 AM EDT up reply actions

That’s a good point. I guess I was trying too hard to eliminate the “park factor” portion of luck, but that would also take out some of the just pure luck. When I have time (i.e. when I finally finish summer classes), I’ll look at a comparison between doing it with ERA-FIP versus ERA-xFIP.

Honor is no substitute for victory.

by The Dark on Jul 15, 2010 8:54 AM EDT up reply actions

You definitely would want to use xFIP or SIERA if you were looking at luck alone. There are three basic kinds of pitching luck:

(1) BABIP
(2) HR/FB
(3) Timing

BABIP is only about 12% the pitcher’s fault, so FIP and xFIP assume a league average distribution. SIERA (full disclosure: my co-invention at Baseball Prospectus w/ Eric Seidman) does take into account BABIP luck because BABIP skill is correlated with K% and GB/FB rates.

HR/FB fluctuations affects a pitcher’s ERA nearly as much as BABIP fluctuations, but after you adjust for park, it is even less the responsible of pitching than BABIP. FIP treats HR/9 as a skill but much of it is luck. FIP is good for removing defense and timing luck, but NOT HR/FB luck. A better measure of luck is ERA-SIERA or ERA-xFIP.

Timing luck is removed from all of them, because it assumes a typical distribution of events. Cole Hamels has made up for his HR/FB bad luck this year with good luck in timing (and neutral BABIP luck). Cole has given up far more HR than his fly balls suggest, but he has been fortunate that those have come with bases empty, so it hasn’t killed his ERA.

Hamels is a great example of a pitcher who will look fake lucky by FIP. By neutralizing his timing and BABIP luck away (good, neutral), but not neutralizing his HR/FB luck (bad), it makes him look lucky. But his SIERA and xFIP both reveal he’s had net neutral luck in 2010.

by Matt Swartz on Jul 19, 2010 2:27 PM EDT up reply actions

Is there any way to statistically examine the effects of timing luck on a pitcher’s ERA? While I have not thought of it before, nights like tonight when the Phillies were able to manage 13 hits but only 3 runs, coupled with reading what you just said, make me wonder how big an effect bad timing can have on pitching metrics.

by dannijd on Jul 20, 2010 1:28 AM EDT up reply actions