Bunts and RE24

This week’s piece at USA Today Sports Weekly mercifully wrapped up my series on the effectiveness of bunting.  In the comments of past bunt-related screeds I’ve shared in this space, mosc has suggested that I look at this year’s bunting data through the lens of RE24, rather than WPA.  I did so for USA Today, and I’ll share an edited version below.

On August 5, the Phillies and Astros were locked in a 15-inning marathon.  Grady Sizemore led off the bottom of the 15th with a single.  Light-hitting Ben Revere laid down a bunt, moving Sizemore to second.  An out and an intentional walk later, Sizemore scored the winning run on Ryan Howard’s single.

Research has told us repeatedly that the sacrifice bunt reduces the number of runs a team can be expected to score, as a big inning rarely materializes with a free out in the middle.

But in situations in which one run wins the game, is bunting a runner into scoring position such a bad idea?  A statistic called RE24 can help us answer that question.  RE24 attempts to assess situational value a player adds by calculating the difference in run expectancy between the start and the end of a given play.

Through the first three months of 2014, National League teams attempted 958 bunts.  Of those, 668 came in traditional sacrifice situations- runners on first, second, or both, and fewer than two outs.  In five of those six situations, bunts have yielded fewer runs than swinging away.

The following table shows the number of runs a team can be expected to score in each of the sacrifice situations (per Baseball Prospectus) and how many runs teams have scored this year when bunting in those situations.

Base-Out Situation Run Expectancy # Bunts Average Runs Following Bunts Bunting Advantage
0 out, 1st 0.827 264 0.754 -0.073
0 out, 2nd 1.044 77 1.130 +0.086
0 out, 1st & 2nd 1.410 83 1.072 -0.338
1 out, 1st 0.478 174 0.385 -0.093
1 out, 2nd 0.622 7 0.286 -0.336
1 out, 1st & 2nd 0.859 63 0.698 -0.161


As evidenced by the table above, bunting a runner to second, whether with no outs or one, is typically a bad idea.  Teams can expect to score about .827 runs with a runner on first and nobody out.  A successful sacrifice bunt yields a one-out, runner-on-second situation, which reduces run expectancy to .62.  Because not every sac bunt results in a base and an out, teams have been a little more fortunate when bunting this season, averaging .754 runs in those innings, still fewer than they score when swinging away.

Bunting a runner to third with no outs has been a more fruitful strategy.  While run expectancy general decreases from 1.044 to .880 when a runner moves to third on an out, such bunts have wrought enough havoc in the N.L. this season that teams are actually scoring 1.130 runs an inning when bunting in those situations.  A sample of 77 bunts may not be particularly illustrative, but the chance for a sacrifice fly or a shallow outfield single certainly adds some benefit to this play.

When Ben Revere came to bat last Tuesday, it made no difference to the Phillies whether they scored one run or several, as one run would clear the dugout in celebration.  In this case, Manager Ryne Sandberg was more concerned with the odds that the Phillies could score one run than with the chance of a big inning.  The following table shows the likelihood of a team scoring at least one run in various scenarios (per Tom Tango’s study of prior years’ results) and how teams have fared this year.

Base-Out Situation Odds of scoring # Bunts Times Scoring 1+ Runs Bunting Advantage
0 out, 1st 0.426 264 0.413 -0.013
0 out, 2nd 0.623 77 0.649 +0.026
0 out, 1st & 2nd 0.632 83 0.482 -0.150
1 out, 1st 0.269 174 0.236 -0.033
1 out, 2nd 0.411 7 0.286 -0.125
1 out, 1st & 2nd 0.421 63 0.365 -0.056


These results still do not recommend the Phillies’ strategy.  While the results are a little closer to breakeven, the only scenario in which bunting has been advantageous is again a runner on second and no outs.  Howard’s well-timed hit may have done the job, but before Revere came to bat, the Phillies’ odds of scoring a run (setting aside Revere’s hitting skills) were about one percent better if Revere had swung away.

If we limit the study above to the 80 times this year that teams have bunted in a sacrifice situation in the ninth inning or later, the results are not much different.  Bunting a runner to second with no outs has reduced run expectancy from .426 to .333.

Of course, teams don’t only bunt in sacrifice situations.  There is one scenario in which teams have been far more successful when bunting this year than when swinging away.  A team in 2014 starts an inning with a run expectancy of .460 runs.  When bunting to lead off an inning, teams are averaging .563 runs, a swing of more than ten percent.  This is skewed by great bunters like Dee Gordon and Denard Span leading off a lot of innings, but it makes sense, as players who have bunted with no one on base this year have reached by hit or error 41% of the time, far better than those who have swung away.

So the conclusion is essentially the same whether one uses WPA or RE24 to evaluate the efficacy of bunting, with perhaps one exception.  In late-and-close situations, bunting runners from first and second to second and third with no outs yields a positive WPA.  RE24 agrees, as the odds of scoring at least one run improve from .632 to .678 when the first out moves two runners up (though total run expectancy decreases).  In practice, though, teams have not employed this tactic well this season.  In 83 such cases, a run has been scored just 40 times, or 48 percent.  Fourteen bunters have hit into forceouts in this situation and another hit into a double play. Thirteen of those fifteen innings ended without a run scoring.

This speaks to the difficulty of bunting a runner to third.  If the pitcher or catcher fields a bunt, the throw to third is shorter than the throw to second, making it harder for the runner from second to reach third before the ball.  Logic would dictate that the same would be true of bunts with only a runner on second, but it seems that the third baseman having to apply a tag makes a lot of difference.  NL teams have employed 77 bunts this year with a runner on second and no outs.  No runner has been thrown out at third on these plays, while four failed fielder’s choices have left runners on first and third with no outs and another thirteen bunters have reached via hit or error.

It appears that bunting a runner from second to third works unless there’s also a runner on first.  In that case, swing away and hope for a crooked number.


Leave a Reply

13 Comments on "Bunts and RE24"

Notify of
Sort by:   newest | oldest | most voted

Why would a player like Revere ever lay down a sacrifice when a drag bunt would likely move the runner over w/a good chance of not giving up an out?

We have to be really, really careful about reaching bunting strategy conclusions based on either Win Probability numbers or RE24 numbers. The problem is that the win expectancy and run expectancy numbers at b-ref and elsewhere assume a league average hitter is at the plate, but most actual batters are either above average or below average, and the best strategy decisions need to be made based on those actual batters. Brian, you point out that that teams can expect to score .827 runs with a runner on first and nobody out. And that the successful sacrifice in that situation cuts… Read more »
I think the point is to look more generally. RE24 and WPA are both flawed on that but I think combined you get a pretty good idea. RE24 doesn’t care about 1 run, it cares about the expected total run return. Similarly, WPA talks about winning the game, fairly useless in early innings where the goal is simply to create runs. I agree with Brian’s conclusions here. Bunting is over-used in a number of situations it should not be used but is very much still a necessary part of baseball. In fact, there are many baseball players who would greatly… Read more »