• Has Joba Been Somewhat Lucky?

    Posted by on July 31st, 2008 · Comments (22)

    One of the stats that they track at The Hardball Times is “Expected Fielding Independent Pitching” (xFIP). It’s a stat that is supposed to a good predicter of a pitcher’s future ERA. It’s sort of like Component ERA.

    For example, last season, Mike Mussina’s ERA was 5.15 – but, his xFIP was 4.58 – and this suggested that Moose was not as bad as he appeared to be in 2007…and could have had a better ERA with some luck/good fielding. (Click here to see where Moose was last year in terms of Component ERA.)

    This season, to date, Joba Chamberlain has an ERA of 2.24 – which is awesome. However, so far, Joba’s xFIP is 3.14 – which is still a great ERA-ish number for a starter to have these days…but it’s not like having one that’s near two.

    Just some food for thought…when looking at Chamberlain’s ERA. It appears that it may be so low via some help from his defense…when you look at his xFIP.

    As a side note, to date, Mike Mussina has an ERA of 3.56 this season versus an xFIP of 3.62…

    …looks like Moose’s ERA is right where it should be, this year.

    Comments on Has Joba Been Somewhat Lucky?

    1. danielwhite2710
      July 31st, 2008 | 12:07 pm

      I understand the reasoning…but I would like to see further research on the xFIP.

      My theory, which would have to be tested, is that pitchers with better stuff, Beckett, Sabathia, Sheets, Harden, etc with always have lower xFIP then league average.

      The reason being that their stuff produces less amount of hard-hit balls making fielding plays more likely.

      Totally a theory, but I think investigation should be done on that before we go ahead and say he has just been lucky…I see the point…but I am wary at this point.

    2. OnceIWasAYankeeFan
      July 31st, 2008 | 1:40 pm

      Well it seems to make sense because for a period of time he was walking people, and he’s also pitched out of some tough situations. I don’t know how the stat is calculated but if his BABIP is lower than expected, I’d expect this ERA measure would be higher than his actual ERA currently is.

    3. DJ21996
      July 31st, 2008 | 1:56 pm

      There is no way that you could watch all of Joba’s starts and even think that he has been lucky.

    4. DJ21996
      July 31st, 2008 | 2:01 pm

      BTW, Joba’s xFIP would be 2nd in the AL behind Halladay if he qualified.

    5. July 31st, 2008 | 2:08 pm

      Whoa – I didn’t say he was lucky…I suggested that maybe he was luck to have an ERA near two. That’s all.

    6. July 31st, 2008 | 2:12 pm

      danielwhite2710, FWIW, some 2007 stats…

      Beckett, 3.27 ERA, 3.56 xFIP
      Santana, 3.33 ERA, 3.55 xFIP
      Sabathia, 3.21 ERA, 3.63 xFIP
      Haren, 3.07 ERA, 3.99 xFIP

    7. DJ21996
      July 31st, 2008 | 2:15 pm

      Too late…you have got me stat hunting now.

      Joba has the 2nd best k per 9 inning ratio among all starters with atleast 50 innings pitched.

      He has the 2nd best FIP of all starters will atleast 50 innings pitched.

      The best home runs per 9 inning rate amongst that category.

    8. DJ21996
      July 31st, 2008 | 2:17 pm

      Striking out a ton + keeping the ball in the ball park = Success.

    9. DJ21996
      July 31st, 2008 | 2:18 pm

      Infact, we have 2 of the top 3 pitchers in keeping the ball in the ballpark.

      Wang and Joba…the dynamic duo. One of them kills you and the other makes you run out grounders only to carry on running to the dugout.

    10. AndrewYF
      July 31st, 2008 | 2:21 pm

      Also, look at his Groundball/Flyball/Linedrive rates. His groundball rate is above 50%, and his line drive rate is a ridiculously good ~12%. Combined with his strikeouts (and, in his past four starts, a near-complete lack of walks), that is why he’s been pitching as well as anyone can pitch.

    11. July 31st, 2008 | 2:59 pm

      As a side note, to date, Mike Mussina has an ERA of 3.56 this season versus an xFIP of 3.62…

      …looks like Moose’s ERA is right where it should be, this year.

      ————-

      Or, is it really the case that Mussina’s xFIP is right where it should be.

      In other words, what is the correct number to look at.

      The proponents of stats like component ERA and xFIP claim that the stat does a better job of predicting future ERA than ERA did.

      So, I decided to put it to the test.

      I was unable to find the exact formula for xFIP, so I could not use it for this test. So, instead I am using component ERA.

      If it is the case that component ERA does a better job of predicting a future ERA than ERA itself, than the following should be the case–if ERA and CERA disagree on a pitcher, then the next year, the pitcher’s ERA should be closer to what CERA said it should be than what ERA said it should be.

      So, I decided to set a test in which I looked at every single season in which a pitcher (1) pitched 100+ innings and (2) had his ERA and CERA disagree by at least 0.50.

      Then, I looked at how often ERA was the better predictor or CERA was the better predictor.

      First, my study only focused on active pitchers. The data there was 40 cases, 19 in which ERA did a better job of predicting what the next year’s ERA would be, 21 cases in which CERA. But, that’s really too close to call. All it would take was just a single case to flip and it’s a 50-50 tie.

      So, I extended the study to focus on every pitcher in baseball history.

      In that study, ERA did a better job 183 times, CERA did a better job 175 times.

    12. July 31st, 2008 | 3:03 pm

      I’m not sure I did a good enough job explaining the methodology, so I will give an actual example of how the study worked–

      Livan Hernandez had a 3.98 ERA in 2005, with a 4.53 CERA, a difference of 0.55.

      In 2006, his ERA was 4.83. 4.83 was closer to 4.53 than 3.98, so CERA was closer to predicting his subsequent season, so it was a victory for that stat.

    13. July 31st, 2008 | 3:03 pm

      1 other thing–in addition to pitching 100+ innings in the season in which the pitcher’s ERA and CERA disagreed by at least .50, he also had to pitch at least 100 innings the next year.

    14. July 31st, 2008 | 5:34 pm

      Maybe the delta of .5 is too small. How would this break out if you used a delta of .75+, or, one or greater?

    15. July 31st, 2008 | 5:49 pm

      Using 0.75+

      ERA: 64
      CERA: 53

      Using 1.00+
      ERA: 23
      CERA: 16

      Based on these figures, it looks like ERA’s winning percentage over CERA increases as the delta increases.

      It went from 51.1% when the cutoff was .50, up to 54.7% when it was .75, up to 59.0% when it was 1.00.

    16. danielwhite2710
      July 31st, 2008 | 6:16 pm

      Ah, that is what I was looking for. Thanks Steve!

    17. July 31st, 2008 | 8:34 pm

      No sweat, danielwhite2710.

      Lee, could park factors be an issue here – with the ERA marks?

    18. July 31st, 2008 | 8:48 pm

      Steve,

      I don’t believe that would have any effects in terms of this topic.

    19. studes
      August 1st, 2008 | 12:01 am

      Steve asked me to comment, but I don’t really have much to add. Several times I have studied ERA vs. component ERA as predictors — here’s an example:

      http://www.hardballtimes.com/main/article/ten-things-i-didnt-know-a-while-ago/

      …and have generally found that no stat is very good at predicting future ERA. Too much random variation.

      But to clear up one thing: Joba’s xFIP isn’t high because he allowed less in-play hits on batted balls. His DER is actually low (which is often the case for ground ball pitchers).

      It’s because he hasn’t allowed many home runs. His ERA is equal to his FIP and the only difference between FIP and xFIP is the home run.

      He’s allowed only 5% of outfield flies to be home runs. That’s a very low figure, but it’s also what he achieved last year. Can he continue on that pace? I personally would be against it, but something in the 8% to 9% range (which is still below average) might be appropriate for him.

    20. August 1st, 2008 | 8:49 am

      Thanks Dave!

    21. Raf
      August 1st, 2008 | 11:37 am

      Thanks Dave!
      ———
      Yeah, thanks Dave, Steve, Lee, and everyone else. Very informative thread.

    22. August 1st, 2008 | 1:42 pm

      I just discovered a typo in my spreadsheet. This doesn’t change the conclusion from my study, but it does change the numbers involved.

      I have seen plenty of times when the proponents of CERA give a list of all of times when a pitcher’s CERA was closer to his next year’s ERA than ERA was and they are clearly giving the impression that CERA is a far better predictor. In fact, if you take the lists at face value, CERA gets it right and they never give any examples of ERA getting it right.

      I wish that they would be honest, like Dave is, and admit that the data does not support CERA being a better predictor. I wish they would include a statement that they are just presenting cherry picked data.

      Looking at the cherry picked lists made me think that those one sided lists give a lot more examples of CERA successes than the average number of successes per year that my side yielded.

      So, I just took another look at my spreadsheet and I see the problem.

      Instead of requiring the pitcher to also have 100+ IP in the subsequent year, I entered the field into Excel and instead the requirement was 100+ ER in the first year. That obviously was not what I wanted and it produced the wrong list of test cases,

      That changes the results of the study, but not the conclusion.

      The correct figures are now–

      DIFF OF .50+
      ERA: 900
      CERA: 888

      DIFF OF .75+
      ERA: 302
      CERA: 278

      DIFF OF 1.00+
      ERA: 86
      CERA: 86

    Leave a reply

    You must be logged in to post a comment.