We're barely a week into the season, and already announcers are mentioning the bunt. I was watching a game earlier this week (I honestly can't tell you which. I'm thinking it was Braves-Brewers or Nationals-Mets), and the pitcher came up to bat with a runner on base and less than two outs. The announcer promptly said, ``And up comes [The Pitcher], and he should be bunting here."
As I stated in the summary, bunting gets a lot of criticism within the sabermetric community. Manny Acta stated at SABR Analytics that he first started looking at sabermetrics after being informed that the inefficiency of the bunt by a Mets clubhouse kid in 2005. So the knowledge about the bunt has been working its way down the line over the years.
That's what you'd think, but before we assume that, let's dig into the data a little. To start, let's look at the RE24 of a plate appearance where the pitcher executes a perfect sacrifice bunt; that is, each runner advances one base and the pitcher is thrown out at first.
|Base State||No Outs||One Out||Two Outs|
|0 0 0||-0.2183||-0.1571||-0.0918|
|0 0 1||-0.0344||0.1979||-0.3527|
|0 1 0||-0.1560||-0.2843||-0.3054|
|0 1 1||0.0279||0.0707||-0.5663|
|1 0 0||-0.1892||-0.1880||-0.2064|
|1 0 1||-0.0053||0.1670||-0.4673|
|1 1 0||-0.1269||-0.3152||-0.4200|
|1 1 1||0.0590||0.0398||-0.6809|
So, there are 6 out of a possible 24 Base-Out states where a perfectly executed sacrifice bunt results in a positive RE24. However, it's not enough to look at the RE24 for the plate appearance, we also need to look at the proportion of times a Base-Out state occurs given that the player bunted, or P( Base-Out | Bunt Occured ). Since we're focusing on pitchers, we're just looking at pitcher bunts.
|Base State||No Outs||One Out||Two Outs|
|0 0 0||0.01530612||0.005102041||0.005102041|
|0 0 1||0.00000000||0.006377551||0.000000000|
|0 1 0||0.06887755||0.006377551||0.001275510|
|0 1 1||0.00127551||0.000000000||0.001275510|
|1 0 0||0.27933673||0.327806122||0.003826531|
|1 0 1||0.02551020||0.048469388||0.000000000|
|1 1 0||0.07142857||0.128826531||0.003826531|
|1 1 1||0.00000000||0.000000000||0.000000000|
So, from all this, we can get an expected value of RE24 given that a bunt occurred but taking ΣBase State ΣOuts RE24|Base-Out × P( Base-Out | Bunt Occured ). After all that, the RE24 for a bunt would be expected to be -0.1755696.
So, if a pitcher can have a better RE24/PA than -0.1755696, his team's run expectancy would be better served on average by him hitting, as counter-intuitive as that might seem. So what percentage of pitchers would you expect to be better off bunting? Of the 75 pitchers who got at least 20 non-bunt PAs over the 2013 season, there were 12 pitchers who fell below the break-even point. That still means a majority of pitches (84%) would still be better off swinging away. For those wondering, there were no position players with at least 400 PAs to fall below the threshold (The lowest value was -0.0624).
So who were the pitchers? Who were the disastrous dozen? Well, here they are, along with some of their other batting stats.
|Pitcher||Non-Bunt PA||RE24||RE24/PA||Slash Line||wOBA||wRC+|
|Jorge de la Rosa||50||-11.3249||-0.226498||.038/.074/.038||0.058||-95|
Blech. I mean really, blech. Those are some rough statistics. Only 3 of those 12 had an ISO better than .000 and there were 4 OBPs over .100. By comparison, Miguel Cabrera had the highest RE24/PA in the majors last year with 0.1153. And for those interested, of the 75 pitchers with 20+ non-bunt PAs, there were only 4 pitchers with a RE24/PA greater than 0. They were Tyler Chatwood (0.0739 in 39 non-bunt PAs), Zach Grienke (0.0308 in 70), Henderson Alvarez (0.0274 in 30), and Scott Feldman (0.0237 in 32).
Of course, there's the obvious cavaet in this: sample size. There's an incredibly small sample size in this dataset, making any conclusions difficult to draw. However, with the well-documented ineptitude of pitcher hitting, these conclusions might not be too big of a stretch, even with the small sample size.