OK, I get it. Nobody really wants to see another discussion of the value of the sacrifice bunt crop up on this site, but I'm going to try to stir one up anyway.
Recently, it has been argued that it doesn't make sense to merely argue about the value of a sacrifice by using Run Expectation tables, because they lose a lot of context. Bunts are highly situational, the argument goes, and examining what happens, for example, in generic situations when there's a runner on first and nobody out versus what happens in generic situations where there's a runner on second and one out misses the point behind the bunt. The game may be very close in a late game and you're only trying to get one run across. The game may be in the early innings and the pitcher is up. Surely it's OK to bunt with a pitcher?
The argument has merits. The aggregated RE tables remove the situational context behind the runners, innings, score, and hitter and it's those situations which precisely drive the decision to bunt or to not bunt.
Then the argument turns to speculation. Isn't it possible that teams that bunt actually BEAT the RE averages because they not only clear the weaker hitters and productively advance base runners, but they do it with the knowledge of the strength of the on-deck hitter. That is, you are creating the runner-on-second-one-out situation (for example), but you are also setting it up so that a better-than-average hitter(s) will be coming to the plate behind them.
This is, of course, a distinct possibility. One that's worthy of examination. And it's the point of this post.
I churned through the 2012 Nationals home games using Retrosheet (*) play-by-play data, and found the scoring results of innings in which a team had a "sacrifice bunt situation", comparing the times when a sacrifice was successfully executed and when one was not. For my purposes, I considered a situation to be a bunting situation if the score of the game was tied or no more than 3 runs to the favor of any team and
- There was at least one runner on base, but the bases were not loaded, and there was nobody out, or
- There was at least one runner on base, but the bases were not loaded, there was one out, and the pitcher was at the plate.
In 23 of the 50 situations in which a sacrifice was executed, the team ended up scoring in the inning (46% of the time). In the remaining 444 situations (i.e. sacrifice bunt situations in which a sacrifice was not executed), the team ended up scoring 47.7% of the time. This represents a small, insignificant difference in the likelihood of scoring following a sacrifice.
When examining the totals runs scored, teams that successfully executed the sacrifice bunt averaged .78 runs in those innings. Teams that opted to not bunt (or were unsuccessful in their bunt attempts) ended up scoring an average of .89 runs in the inning.
The distributions of runs scored look like this:
As you can tell, the distributions are almost identical, with the tiniest-seeming greater likelihood of scoring zero or one run when teams bunt, and a similarly tiny greater chance of scoring more than one run when the team does not bunt. This is pretty much exactly what I would have expected the answer to be.
Conclusions: For the most part, the conclusions will be inconclusive. There's a suggestion that bunts may possibly cost some scoring opportunities and that the runs scored when bunting may be slightly reduced, but in neither case are these suggestions demonstrated with any statistical significance. It is possible that pursuing this line of inquiry with a greater amount of data would provide that significance, but frankly I'm too lazy to do that at this time.
I have long argued that sacrifices are not the panacea that the FP Santagelo's of the world would have you believe, but neither are they the anathema some on here see them as. They are, I believe, a mostly neutral, perhaps just slightly less-than-optimal strategy. I'm not fond of them but I don't hate them.
(*) Retrosheet: http://www.retrosheet.org/game.htm I used only the Nats 2012 home season because those were the ones I had previously downloaded. It's kind of a pain to go through and parse the data for all the games for all the teams, ... so I didn't.