Getting Friendly with Pitcher K%

This sunset has a pretty high Glory Factor (GF).

The attentive reader will note that FanGraphs Dark Overlord David Appelman has recently descended from atop his inlaid throne and made some changes to the way that strikeouts are presented in these pages. For batters now, strikeout percentage (K%) uses plate appearances, and not at-bats, in the denominator. While this creates pretty substantial (ca. 6%) movement for Three True Outcome-ist Adam Dunn, most players exhibit a fairly uniform decrease of about 2%, meaning that readers are able to mentally convert to the new format with some ease.

The addition of strikeout percentage (also K%) for pitchers, however — while glorious as the sunset over Quetena Chico — makes for a slightly more challenging mental covnersion, as readers will undoubtedly have become quite accustomed to assessing pitchers using strikeouts per nine innings (K/9).

As Mr. Matt Swartz demonstrated in these pages yesterday, however, knowing a pitcher’s K% is important for at least two reasons (in that it’s predictive of BABIP allowed and home runs per fly ball). Beyond that, K% is plainly the more accurate representation of what we’re intending when we cite K/9 — namely, to note which pitcher or pitchers are the best at inducing strikeouts.

To aid the readership in its transition from K/9 to K%, below are some notes on the two metrics and their relaionship with each other, including some illustrations of K/9’s weaknesses.

1. K/9 and K%: A Stong Correlation
It probably won’t come as a shock to learn that K/9 and K% correlate strongly. Here, for example, is a graph looking at the relationship between K/9 and K% for all 257 qualified starting pitchers between 2008 and’10.

As you can see, the correlation is something in the vicinity of 0.99. While there are certainly some outliers, the conversion between K% and K/9 is a pretty regular one.

2. Converting Between K% and K/9
Because the correlation between K/9 and K% is so stong, the following quick conversion table (using the formula from the above graph) will serve you well in almost every case (K/9 is expected K/9 given K%):

3. Some Benchmarks for K%
While readers will no doubt have an intuitive sense of what constitutes a good (and not-so-good) K/9, it’s likely that K% is a trickier subject. For your edification, then, below is a summary of K% in context (à la mode of House Librarian Steve Slowinski), using number from 2011’s 106 qualified pitchers.

4. A Look at Some Outliers
Though predictive in most cases, the formula to convert K% to “expected” K/9 (or, xK/9 below) produces some outliers. These outliers are actually helpful, as they call attention to those aspects of strikeout-inducing that K/9 misses, but which K% identifies accurately.

Firstly, here are the 10 qualified pitchers from 2008 to ’10 whose K/9 most underrated strikeout ability:

And now, here are the 10 qualifiers whose K/9 most overrated strikeout ability:

5. A Look at Some Outliers, Part II
Looking at the two charts above, you probably already get some sense of the differences between the pitchers who’re under- and overrated by K/9. For example, Roy Halladay and Cliff Lee and Zack Greinke are really good. Ian Snell, on the other hand, is more of a “minor leaguer.”

In any case, to make the relationship clear, here are the averages for each set of 10 pitchers in some notable categories:

Underrated Pitchers

BB/9: 1.77
BB%: 4.98%
BABIP: .274
xFIP-: 79

Overrated Pitchers

BB/9: 3.65
BB%: 9.11%
BABIP: .323
xFIP-: 96

As you can see, the overrated group was actually rewarded with higher K/9s because they extended innings through walks and hits on batted balls. While the latter category is likely part bad luck, the former is certainly not. The advantage of K% is that it represents precisely how many batters are being struck out and rewards efficiency.

Image courtesy Jeronimo’s Eye.

Print This Post

Carson Cistulli has just published a book of aphorisms called Spirited Ejaculations of a New Enthusiast.

17 Responses to “Getting Friendly with Pitcher K%”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. Kirkwood says:

    Well hey there, Pitcher K%, can I buy you a drink?

    Vote -1 Vote +1

  2. MikeS says:

    So if Adam Dunn were a pitcher his k/9 would be downright Marmol-ian?

    Vote -1 Vote +1

    • Anon21 says:

      This particular season, it would be more appropriate to call it Romorean (?) or Kimbrelian.

      Vote -1 Vote +1

  3. JWTP says:

    Is there a link to find data for correlations of K% with BABIP and HR/FB?

    Vote -1 Vote +1

    • James says:

      There are the SIERA articles (part 1 linked above), and I believe you can always export the values from the Fangraphs leaderboards to give yourself unlimited freedom to search for correlations.

      Vote -1 Vote +1

  4. RC says:

    “As you can see, the overrated group was actually rewarded with higher K/9s because they extended innings through walks and hits on batted balls.”

    Something we’ve been saying for months. Now if only we could fix the same problem in FIP/xFIP….

    Vote -1 Vote +1

    • Deadpool says:

      Wouldn’t take much, just use the expected K/9 a shown here to generate new weights for a K. I do think though that the outliers on either side are going to keep the formula almost exactly the same, and with a 99% correlation the number of significant outliers are probably too small to move the equation even if they don’t wash out.

      Vote -1 Vote +1

    • Deadpool says:

      Unless your suggesting setting FIP against BF?

      Vote -1 Vote +1

      • RC says:

        I’m suggesting that the entirety of FIP is broken for the same reason that K/9 is a broken statistic.

        70% of the events that FIP measures are actually Balls in Play, because it uses IP as a denominator, just like K/9.

        Essentially, the further from average the pitcher’s BABIP is, the more of an outlier he is going to appear in the k/9->k% and BB/9 ->BB% conversions. And the more of an outlier hes going to be between FIP and whatever comes next.

        Vote -1 Vote +1

      • Ben Hall says:

        Saying it’s broken undermines your credibility. Your points are good ones, but neither FIP nor K/9 are broken. Given that you want to replace K/9 with K% and yet the results are almost indistinguishable, K/9 obviously works fine.

        Yes, they can be improved. But when you go overboard nobody is going to listen to you.

        Vote -1 Vote +1

  5. CheeseWhiz says:

    Thank you so much! I’ve been asking for pitcher K% around here for months specifically because K/9 punishes efficient pitchers (Lee, Halladay) and benefits pitchers who see a lot of ABs due to walks and hits.

    Vote -1 Vote +1

  6. Max says:

    So does this help explain the difference in how Brandon Morrow is viewed by Dave Cameron vs. everyone else?

    Also, maybe you could put xK/9 on the player pages too?

    Vote -1 Vote +1

  7. Aren’t the two stats measuring different things, both of which useful? K%’s benefits are listed above, but K/9 just measures the percentage of a pitcher’s outs that are strikeouts. Rather than telling you what percentage of batters faced result in K’s, it’s telling you the percentage of outs resulting in K’s.

    Don’t GB% and FB% use IP as a denominator? Those show percentage of [GB or FB] out of outs recorded, so why use a different denominator for K?

    Vote -1 Vote +1

    • DD says:

      Better yet, why not just be consistent and use GB/9 and FB/9? Same denominator, same naming convention.

      Vote -1 Vote +1