Announcement

Collapse
No announcement yet.

K/TBF better than K/9?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • K/TBF better than K/9?

    Interesting article here... The author suggests that K/9 is not the best measure of a pitcher's strikeout ability, and that we should use K/TBF instead. The rationale is that two pitchers can have the same K/9, but if they did it in a different # of TBF (i.e. one pitcher gave up more hits and/or walks), the one who did it while facing fewer batters is actually the better K pitcher. On one hand, this makes sense; the guy with the higher K/TBF is clearly striking people out more efficiently. On the other hand, the difference between a pitcher's K/9 & K/TBF is based on the number of non-K events (hits, walks) he allows -- and we know walks speak more to "control" than to "stuff", and that pitchers have little ability to control hits on balls in play -- so K/TBF may in fact do a worse job of isolating "K ability" than K/9. Thoughts?
    Last edited by Davis21wylie; 03-27-2008, 09:02 AM.
    World Champions:
    2007 • 2004 • 1918 • 1916 • 1915 • 1912 • 1903

  • #2
    I did not read the article, but I have always felt that K/PA (or BF or whatever) is a truer indication of "strikeout ability", although that is sort of a nebulous term. I like to look at K/PA as a success/opportunity ratio (with a strikeout being a success and any non-K being a failure) and K/IP as the percentage of outs recorded on Ks. Or, K/PA as the nebulous "ability" and K/IP as "reliance".

    The point about hits on balls in play seems to be a point in favor of K/PA, not against it. If a ball is hit into play, then regardless of whether it is a hit or not, it has the same effect on K/PA...it lowers it, as the event counts as a failure. However, a ball that falls in for a hit will have no effect on K/IP, but one that is caught will be lower it.

    If one was a true-blue DIPS believer (I'm not suggesting that anyone is, should be, or that DIPS requires one to be; just using the most extreme position one could take), then K/PA would be a no-brainer. And as a matter of fact, the original Voros DIPS holds K/PA constant between a pitcher's actual line and his DIPS line.
    Last edited by Patriot; 03-27-2008, 04:55 PM.

    Comment


    • #3
      OK...minor rant.

      Pitching is about opportunity, probability, and skills that alter the probabilities.

      Every single plate appearance is an opportunity.

      You must start by accounting for three kinds of opportunities however.

      Opportunities to create a DIPS event: PA
      Opportunities to allow a hit on a ball in play: BIP
      Opportunities to allow a baserunner to advance in some way other than by batting skill: # of Baserunners

      Once you have those things...you take the baserunner events and normalize for how the team as a whole did at preventing baserunner advances relative to the league average and recalculate the top two items based on the number of outs and therefore lost plate apperaances and balls in play you could expect to be added.

      Then you take your BIP data and normalize for how the team defense did at preventing hits on balls in play relative to how the league did, recalculating the number of PA based on the outs added by normalizing the BIP results.

      Then you carry forward the pitcher's DIPS rates per PA (HRs adjusted for park effects) and put the line back together again.

      Comment


      • #4
        K/9 is more useful because it is a much faster calculation and the additional "correctness" brought to the party by K/TBF isn't worth the aggravation to acquire it. Heck, you could even argue that K/AB would be more accurate yet. Why include PA's in which the batter didn't have a chance to strike out??
        Buck O'Neil: The Monarch of Baseball

        Comment


        • #5
          You mean when the count was 3-2 and he walked, he didn't have a chance to strike out?

          Or, when he put the ball in play on a 2-0 count, he did?

          In any case, once pitch-by-pitch becomes more commonplace, we'll be looking at swing&miss, called strikes, and so on.
          Author of THE BOOK -- Playing The Percentages In Baseball

          Comment


          • #6
            Originally posted by Davis21wylie View Post
            ... and we know walks speak more to "control" than to "stuff", and that pitchers have little ability to control hits on balls in play -- so K/TBF may in fact do a worse job of isolating "K ability" than K/9. Thoughts?
            Walks and "stuff" are related. One way to avoid a walk is to let the batter hit the ball. Pitchers who make batters miss the ball entirely go deeper into the count more often, resulting in more strikeouts and walks. I recall an observation that appeared in an old Baseball Abstract. A Toronto fan had kept detailed pitch-by-pitch records of Blue Jays games. He noted that Jimmy Key and Dave Stieb had nearly identical percentages af ball/strikes thrown. Key was regarded as a control pitcher and Stieb a power pitcher (this was around 1985 or so). Key walked and struck out fewer batters because it was easier to put the ball in play off his pitches.

            The fact that pitchers have little control over the results (hits or outs) on balls in play works in favor of K/BFP, not against it. Poor fielding teams will raise a pitcher's K/9 slightly. A pitcher on a bad fielding team will get more strikeouts than one with a great defense behind him simply because he will end up facing more batters per inning.

            Comment


            • #7
              I've always prefereed K/PA.

              I also agree that generally more K's will lead to more BB's, as pitchers are throwing more pitches, getting deeper into the count.

              Back in the early 80's, when I should have been paying more attention to the professors, I doodled a formula, chaining pitch by pitch (assuming the ball/strike% was the same on each count, which it isn't) which would give the expected BB/PA and K/PA based on strikes/pitches (control) and bip/strikes (contact rate). It could also estimate pitches based on BB/PA and K/PA. It also needed foul ball pct, which I didn't know until it recently became available on BBref, and which also showed that there is very little variance between pitchers in foul% when compared to their other pitch rates.
              Baseball Prospectus articles
              FanGraphs articles
              MVN Statistically Speaking articles
              Seam Heads articles

              Comment

              Ad Widget

              Collapse
              Working...
              X