Pitching by the Numbers: Model behavior

  • Oops!
    Something went wrong.
    Please try again later.
  • Oops!
    Something went wrong.
    Please try again later.
In this article:
  • Oops!
    Something went wrong.
    Please try again later.
  • Oops!
    Something went wrong.
    Please try again later.
Collin McHugh is one of many SP sleepers that haven't panned out.  (Getty)
Collin McHugh is one of many SP sleepers that haven't panned out. (Getty)

There’s no sugar-coating it: This year has been a really bad one for the strikeout-minus-walk pitcher prediction model. That’s after two really good seasons for it in 2013 and 2014. 

So what went wrong? Remember, we used the model to highlight the bargains that you could get in double-digit rounds, thus loading up on hitters and power closers in the single-digit rounds. Among the recommendations that were busts for non-injury reasons: Phil Hughes, Colin McHugh, Matt Shoemaker, Ian Kennedy, Drew Hutchison, Shane Greene. There were some hits in there but this is really bad. Most of the cheap recommendations have been brutal.

[Play a Daily Fantasy contest for cash today!]

The big problem with projecting off of last year’s (strikeouts-minus-walk)/IP numbers in March is that the offseason is a reset. You have to accept that there is a good chance that pitchers, especially unproven ones, will not replicate those strikeout and walk numbers and, when they don’t, their entire projection falls apart. 

But if we’re saying strikeouts and walks are unstable year-to-year, what is stable? Certainly not ERA and WHIP, which are largely a function of strikeouts and walks. So by saying we can’t believe strikeouts and walks, we’re really saying we can’t believe anything except for the veteran pitchers who consistently post solid numbers and who are still squarely in the prime of their careers. In other words, we’re saying you have to draft pitchers early — very early. And while some will surely reply in the comments that they have done that and are winning, there is no doubt that loading up on early-round pitchers is the tougher road to winning a league because the surplus value we need to win is much harder to find among middle- and late-round hitters. 

Our ace in the hole with this “wait on pitching” strategy is using the same (K-BB)/IP metric early in the season (since it stabilizes after relatively few batters faced - 126 for Ks and 303 for BBs).   We took an early swing in early May and the pull was disappointing because there weren’t enough outliers on the list.The only contrarian recommendations available on that list were buying Clay Buchholz (6.03 ERA at the time) and picking up Bartolo Colon on waivers. I’ll say that’s a split because Buchholz was very good from that point forward until he got hurt. Colon has been terrible, which we should have seen coming because his entire game is just throwing fastballs, which are great for walk prevention but bad for hit prevention.

We looked at only the last six weeks in the stat about a month later and the results here were winning. The recommendation was to believe Noah Syndergaard and Lance McCullers and it noted four waiver-wire pitchers who should be much better in ERA: Drew Hutchison, Ian Kennedy, Chris Heston and Mike Foltynewicz. Kennedy has been good since then (2.68 ERA, 1.12 WHIP) and so has Heston (2.20, 1.09 but with a sharp decline in K-BB). Hutchison has been terrible mostly because he gave up eight runs the next day. Foltynewicz was soon sent down and came back as as reliever and thus is a non-factor. 

Still, this in no way was enough to allow us to make up ground for those draft-season busts. So the owners who used this model as a tool to help win their leagues in 2013 and 2014 are hating it now. 

In its most extreme application - no starting pitchers selected in double-digit rounds - this model is very volatile, of course. As my friend Gene McCaffrey says, “To be very right, you have to be willing to be very wrong.” Remember, it’s hard to win. In a 12-team league, chalk is about 8% win probability. When you hit on the low-cost pitchers and thus also load up on more highly-drafted hitters, your win probability doubles or even triples. When you miss, though, it’s very difficult to even place.

But you don’t have to play this draft game with your pitchers as extremely as I advocate. You can middle it by taking one ace and one elite closer in the single-digit rounds. But just make sure it’s a true ace and not a pretend one like Jordan Zimmermann or James Shields. True aces require a pick in the first three rounds and probably in the first two. This, to be sure, is going to put you behind the eight-ball with hitters. But you do have more bullets with the hitters to hit on a cheap impact bat. 

The bottom line is that the predictive success of (K-BB)/IP is limited, of course. But that doesn’t mean that it’s STILL not the best tool in the toolbox. You can limit your exposure by minimizing your risk with an early starting pitcher, increasing your odds to cash but lowering your odds to flat-out win. And we all agree it’s more fun to be in the money than to be out of contention before training camp starts. This model worked better than could be expected the prior two years and worse than anyone could fear this year. It’s still been a net positive. So we will keep (K-BB)/IP in the toolbox, hopefully with much better results going forward.