Going to WAR for the Cy Young Award?

As I was putting together my ballot for The Hardball Times postseason awards, I arrived at an intellectual impasse. The NL Cy Young Award has three very strong candidates, and while my instinct was to lean toward Clayton Kershaw, I was tempted to consult the numbers—not the cosmetic, mainstream stats, but the gold ol’ BABIPs, and FIPs. I was tempted to go find out whether one of the worthy candidates benefited more from luck than the others.

But then I was struck with a question—is it appropriate to discount luck for these awards? What should we be rewarding, inherent performance or outcome?

I’ve long since comfortably reconciled the alleged false dichotomy of the word “valuable” in the context of the MVP; value can only have one reasonable meaning in that context: the highest absolute value beyond zero. But, the question of whether to consider what “should have” happened to a pitcher (or position player) introduces a new, more nuanced wrinkle into how one interprets best performance. Considering wins for example, clearly privileges (and inaccurately credits) outcome too far above underlying performance level—but where is the balancing point for what did happen and what should have happened? I know ERA+ should be privileged over wins in this context, but should ERA+ be discounted too, in favor of FIP?

While I’m not completely convinced that I will not change my mind on this, my thought is that what did happen is far more important in the context of determining the best player that what quite possibly should have. Outliers may be outliers, but that doesn’t mean that they didn’t happen. If, for example, that player got quite lucky with his home run/flyball rate, or strand percentage, I’m fine with rewarding that—those events still benefited his team and resulted in win value.

Crediting that is not on the same level as being swayed by W-L records. At some point, we have to recognize what actually happened. The guy who hits 10 improbable roulette bets does after all go home a rich man, even if he is a rich, dumb man.

One other observation I’d like to make here: I do find it somewhat telling that I had no temptation to look up the core indicator stats for position players. Perhaps that has to do with the fact that the choices for best position players in each league seemed clear cut (not to say Justin Verlander shouldn’t receive MVP consideration). Overall, though, this probably speaks to an inherent difference in the default degree of legitimacy we grant semi-advanced offense stats (like OPS+), and the greater scrutiny we feel obligated to apply to pitching stats. And, perhaps there’s some merit to that feeling.

Print Friendly
 Share on Facebook0Tweet about this on Twitter0Share on Google+0Share on Reddit0Email this to someone
« Previous: How dumb are we? (2011 edition)
Next: 40 years later: Short still stinks »


  1. Brad Johnson said...

    I agree that our considerations should focus on what happened rather than what should have happened. For example, if a Cy candidate had a .260 BABIP, we don’t KNOW exactly what his BABIP really should have been. There’s a very outside chance that it should have been .260 or lower. More likely, it’s somewhere between .270 and .300 with another outside chance of being higher than that. We can use xBABIP to narrow it down, but that comes with a sizable error term too. The result is that we have to fall back on our biased interpretations or flawed proxy stats to distinguish the correct number.

    Meanwhile, we know exactly what physically happened. It’s much firmer data to play with.

    Who were your obvious MVP candidates? I grappled with Kemp/Braun in the NL and Bautista/Ellsbury in the AL.

  2. Michael said...

    I have come to step away from the stats-only mentality, not because I don’t love sabremetrics (because I do – I love the knowledge it gives, the learning). However, if we focus too much on the advanced stats, we begin to fail to see the forest for the trees. There is beauty in what happens on the field, and it shouldn’t be ignored because of what we think should have statistically happened. What I am saying is that we need to find a middle ground, where we enjoy and appreciate the now of baseball, while still striving to explain/predict the future without expectation of results.

  3. John C said...

    You vote based on what actually happened, not what your metrics think should have happened. If some guy goes 22-4 with an ERA of 2.15 but had a WHIP of 1.65, he should still get votes for the Cy Young Award because he did deliver the results you wanted. I probably wouldn’t want that pitcher on my team next year, but that’s not relevant to what he did this year.

    Stuff like FIP and WHIP is useful for evaluating pitchers going forward. It shouldn’t be used in trying to evaluate the actual value of what a pitcher has already accomplished. A pitcher with an ERA of 2.15 had an ERA of 2.15, period, and that means he saved a bunch of runs. Even if he was lucky as heck.

  4. Michael said...

    John, I kind of agree with you, and kind of don’t. If you have two equally similar stat lines, then I think it is appropriate to look at predictive stats. I would never say to rely solely on ERA and Wins, but also look at Innings Pitched, WHIP (which is useful evaluating pitcher now), K-rate, BB-rate, and K-BB rate.

  5. deron said...

    Kershaw has won the pitcher’s triple crown; the discussion should end there. But for any doubters, here’s the cincher: compare how the top three candidates Kershaw, Halladay, and Kennedy fared against the top twenty NL hitters and the rest of the league. Against the rest of the league: OPS allowed was .548, .547, and .601 for Ker, Hal, and Ken, respectively. Now, against the 20 hitters Hal’s was .912, Ken’s was .948, and Kershaw… .610! 

    I included Kennedy instead of Lee because of Lee’s 17 wins not being enough to beat Kershaw or Halladay. Although I concede that good arguments can be made to include Lee over Kennedy, it doesn’t change the fact that Kershaw should win.


Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>