Like us on Facebook


Follow us on Twitter





Results 1 to 3 of 3
  1. #1
    Join Date
    Oct 2007
    Posts
    10,600

    Lincecum and Verlander in 2010 Analysis

    I find ERA predictors are often misused, and often tell strange stories. To illustrate my point, let's consider a fake conversation between me and myself about the 2010 seasons of Justin Verlander and Tim Lincecum so far:

    Person A: "Verlander and Lincecum have had roughly equal seasons in terms of productivity because their FIPs are 3.34 and 3.35, respectively."

    Person B: "No you idiot! xFIP has their rates sitting at 4.06 for Verlander and 3.43 for Lincecum! Lincecum has clearly been superior."

    Person A: "No you're wrong because their ERAs agree with their FIP similarities - and after all, ERA is pretty much an actual results measure. (3.72 for Lincecum and 3.65 for Verlander)."

    Person B: "You're an idiot for using ERA as evidence! xFIP normalizes their home run rates, which makes it more accurate because clearly Verlander's 5.9% HR/FB ratio is ripe for regression. After all, he is an extreme flyball pitcher with a GB/FB under 1.00 for his career while Lincecum has typically gotten around 45% GB rate with a 1.35 GB/FB ratio for his career."

    Person A: "So you punish Verlander for home runs he hasn't allowed? He may be a flyball pitcher but his 7.8% HR/FB ratio is good. xFIP is wrong precisely because it normalizes in this case. Doc Halladay is a groundballer owning a 2.19 GB/FB ratio for his career and he still allows a 9.2% HR/FB."

    Person B: "But that's because Halladay is an extreme groundballer, so it stands to reason that when he allows flyballs they aren't lazy flies but hard hit balls, boosting that HR/FB."

    Person A: "So then you agree that HR/FB will vary based upon arsenal? Then you would also agree that xFIP normalizing a key component is probably a faulty assumption, given our discussion."

    Person B: "Well...not in this case because Lincecum isn't an extreme groundballer. SIERA says Verlander has a 3.72 while Lincecum has a 3.38, and they take into account the relative usefulness for HR rates for strikeout pitchers (which they both are)."

    Person A: "That's all well and good but the SIERA fix is based upon the assumption that HR/FB matters less because the strikeout pitcher allows less baserunners, leading to more solo shots. Verlander has a 1.25 WHIP this season while Lincecum has a 1.35 WHIP, so clearly Verlander has also done a better job of not allowing baserunners in addition to not allowing home runs."

    Person C: "You're both wrong. tERA says Lincecum has a 3.81 while Verlander has a 3.34. I would be inclined to agree with Person A, but since tERA is the best measurement of batted ball types, and we're talking about two pitchers with similar strikeout and walk rates, Verlander has clearly had the superior year."

    Persons A and B: "Batted ball data is highly volatile. Attempting to measure 3/4 of a season's data on batted balls is a pointless exercise. Plus, the 30 point drop in BABIP from Lincecum to Verlander means Lincecum is getting very unlucky. You are more wrong than anyone ever."

    Here's the tl;dr version:

    Person A: The years are the same. SIERA and xFIP unfairly punish Verlander because he isn't allowing very many home runs while being a flyball pitcher. The K/BB rates are similar and FIP proves that the pitchers themselves, not their defenses, have had roughly equal years. Not to mention the fact that their ERAs are similar, which means the actual results on the field were very close.

    Person B: Lincecum is better. SIERA and xFIP both agree he's been about half a run better because he strikes out more people, he induces a heftier percentage of grounders to cancel out the .10 difference in WHIP, and he does it all while maintaining a pretty good HR/FB. Verlander's HR/FB is the reason FIP and ERA overvalue him, and it's so low that it's due for regression.

    Person C: Verlander is better. tERA shows he's not allowing hard contact at all. Their FIPs may be similar but the 3% difference in LD% is significant, as is the fact that he's allowing a ton of fly balls and very few are going for homers. As you would expect, Verlander's IFFB% (popup percentage) is a gaudy 11.8% while Lincecum's is only 6.3%. Verlander may not be the Johan Santana of popups yet, but he's excellent at getting weak contact of the other variety while striking out a ton of people.

    Who do you think is right? And why?

  2. #2
    I don't know if there's a "right" way or a "wrong" way about it. Hitters are pretty straightforward to assess; pitchers are far more difficult. I'm not the hugest fan of FIP as it fails to account for batted ball types- but, on the other hand, the biases in batted ball types is so problematic that we can't take those as gospel either.

    No matter how we look at it, there's going to be an issue, somewhere. My personal preference is straight BaseRuns; if I'm looking to make it a defense independent estimator I incorporate simple batted ball types (GB, FB+LD - IFFB) to try and lessen the batted ball problems. And of course there's going to be issues with that, too.

    Since we're on this topic, I'd like to see one of these estimators account for things like inducing double plays or picking off baserunners. I haven't seen a correction for infield hits allowed, either. I don't think this is accounted for in tERA, but I could be wrong.

    And since I'm rambling, I wish we'd stop making these ERA estimators. Just switch it to absolute runs per nine. ERA is just a lazy man's DIPS.
    My blog- analysis of the San Francisco Giants, Baseball, and Sabermetrics.

  3. #3
    Join Date
    Jul 2009
    Posts
    57
    Given that the differences in next year RMSE between SIERA, xFIP, and tRA are very small, and that the two metrics take very different approaches to their treatment of batted ball types, I'd say we are not yet in a position to have a confident answer for pitchers who diverge in the metrics (and that's leaving aside the argument about whether actual results should be considered, in which case you can bring ERA and ERC back into the discussion).

    I prefer to look at five different run averages: actual run average, and four other BsR-based metrics: component RA (taking actual event types allowed--singles, doubles, etc.), straight DIPS RA (obviously a very blunt tool but one that is useful as a comparison since it takes balls in play out of the mix entirely), a tRA type that considers all batted ball types at their actual LW value, and a SIERA type that looks at grounders and non-grounders.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •