Rob Neyer

MLB | GameCast
Scores | RealTime Insider
Pitching Probables
Standings | Grid
Stats | Stat Pack
Injuries: AL | NL
Minor Leagues
MLB en espanol
Message Board

Park Factors
News Wire
Daily Glance
MLB Insider

Player News | Rater
Daily Fantasy Buzz
Out of the Box
ESPN Auctions
Tuesday, August 20
Surprise, surprise: Bunting may be OK

By Rob Neyer

In yesterday's column, I wrote

But does the bunt make sense, ever? Yeah, it can. When you've got a pitcher batting. A weak-hitting pitcher. Otherwise, it rarely makes sense. And I'm basing this contention not on some abstract statistical formula, but on a completely practical application. Tom Tippett, the wizard behind Diamond Mind Baseball, looked at actual play-by-play data to see if sacrifice bunts increased or decreased a team's chances of scoring runs.

He found that sacrifices -- whether successful or not -- actually decrease a team's chances of scoring runs. And further, that a sacrifice usually decreases a team's chance of scoring one run. So why do it?

Well, that's not what Tippett found at all. My sloppiness -- and yes, my predisposed biases against the bunt -- led me to remember what I wanted to remember, rather than what Tom actually said and wrote.

Rather than summarize what Tom really did say -- I'd probably just screw it up again -- I'll let him speak for himself...

I just saw your column, and while I'm very grateful for the mention, I'm afraid it's not what I was trying to say. Here's a short version of what I meant to get across in the talk (obviously without as much success as I'd hoped, since more than a few very smart people thought I said something else):

  • For many years, baseball analysts have used expected runs and one-run probabilities to assess the before and after states for a tactic with two outcomes ... using these values, one can then compute the probability of success that must be achieved in order to justify taking the risk ... I'll call this approach "traditional break-even analysis."

  • When using this approach, analysts tend to use values for expected runs and one-run probabilities that are based on league averages. Using those values, we have to conclude that the bunt almost never makes sense, because the cost of the out is almost always greater than the value of the base gained, and this is true whether you're playing for a big inning or for one run.

    The only exception is when you've got runners on first and second with nobody out, and you try to bunt them over to second and third, in which case your chances of scoring one run go up by a tiny amount

  • Wondering whether this conclusion was biased by the use of league averages, I wrote a program to measure expected runs and one-run probabilities for any sequence of hitters against any pitcher in any ballpark in any era.

  • With these specific values in hand, I figured I would find situations where the bunt did indeed make sense, but I was surprised to find that it didn't work out that way. These values brought us closer to being able to justify bunting in some situations, but still left us short of being able to defend most of the sacrifice attempts that we see in today's game.

    If I had stopped here, my conclusion would have been that bunting doesn't make sense for position players or even for pitchers who aren't hopeless at the plate.

    However, there's another (and much bigger) problem with traditional break-even analysis: it accounts for only two possible outcomes ... a successful bunt is assumed to concede an out for a base, while a failed bunt is assumed to concede an out for nothing

    But my study of real-life play-by-play data showed that there are enough singles and errors on bunt attempts to justify sacrifices in many situations after all. In fact, in the American League last year, teams that bunted after a leadoff single or walk scored just as many runs as teams that didn't bunt in these situations (I looked at the AL so I could focus on bunts by position players).

    As a result, I concluded that the sacrifice bunt is a viable tactic; it's the traditional break-even analysis that is flawed. Partly because that technique doesn't use lineup-specific run values, but mostly because that technique ignores the hits and errors that don't cost you any outs.

  • Tom has the facts and figures to support all of this, but I won't reproduce them here because he's planning to write all of this up in greater detail for the Diamond Mind site.

    The point here is that if you include all possible outcomes of the bunt -- that is, the errors and the hits, along with the various outs -- there are many situations in which the sacrifice attempt is a good move.

    Interestingly enough, Bill James anticipated this conclusion five years ago in his book, The Bill James Guide to Baseball Managers. He explored some of the same issues that Tippett has, and decided that there probably are situations where the bunt makes sense. And James concluded by suggesting, "The rest of us need to keep an open mind."

    I agreed with Bill then, and I agree now. Wouldn't it be a hoot if someday we discover that the 2002 Oakland Athletics failed to win the West because they didn't bunt enough?

    Yesterday in a Quick Hit, I wrote about Tom Glavine never having given up a grand slam -- and yes, I forgot that Jim Palmer faced both Mickey Mantle and Roger Maris -- and Tom Ruane responded with his usual insightful analysis ...


      Interesting comment on Palmer and Glavine's performance with the bases loaded. One thing you might want to look at is the number of plate appearances each pitcher has had in these situations. Here are Glavine's career numbers with the bases loaded:

       IP    AB   H  2B  3B  HR  BB   K  HBP  SF   AVG
      3300  268  63   9   1   0  17  53   1   22  .235

      I don't have the numbers for Palmer's entire career (Retrosheet hasn't released 1965-66 and 1970-73 yet), but for the years I do have, here are the numbers:

       IP    AB   H  2B  3B  HR  BB   K  HBP  SF   AVG
      2490  104  18   3   0   0   7  17   0   11  .173

      The striking thing about this is how rarely Palmer even found himself in a bases-loaded situation. During the years I have, he faced a batter with the bases loaded on 122 occasions, or once every 20.4 innings. Glavine has been in a similar situation 308 times, or once every 10.7 innings. In terms of batting average, Palmer was much tougher than Glavine in these situations, but strictly from the point of view of keeping the ball in the park, I'd warrant that Glavine's achievement is already much more impressive than Palmer's.

      Keep up the good work and talk to you soon.

    For those of you who don't know, Tom regularly engages in some of the most illuminating research out there, thanks to his hard work and his facility with the play-by-play data that Retrosheet has released.

    And finally -- not to open up a can of reactionary worms or anything -- I was both amused and saddened to see the following headline:

    Workhorse Burnett's season may be over

    Hmmm ... "Workhorse" and "season may be over" ... Might there be a connection here?

    And then you click on the link, and you read quotes like:

    "We would never do anything to hurt him. With the future the guy's got? No way." (Marlins manager Jeff Torborg.)


    "We didn't see this coming. He has been a workhorse all year." (Marlins general manager Larry Beinfest.)

    In A.J. Burnett's last five starts, he's thrown 132, 128, 93, 123, and 117 pitches. And I'm keeping an open mind: maybe there's not a connection here.

     More from ESPN...
    Neyer: Making sacrifices
    Most of the teams with few ...
    Rob Neyer home page
    Everything you ever wanted to ...

    Rob Neyer Archive

     ESPN Tools
    Email story
    Most sent
    Print story
    Daily email