Sabermetrics 101: Bunting

The problem is that those statistics represent an average over dozens of years and don't tell us too much on their own. These are the average runs during those situations... on average, with average hitters on average teams in average situations. We don't even know the standard devations. Are 0.78 and0.70 actually even different staistically speaking? 

Furthermore, bunting happens in a situational context when you know that in your particular case the odds are different than average. If I have a hypotehtical situation with Buxton on first with Matt Tolbert at bat... I'm going to bunt every time knowing that there is a chance Buxton ends up on third with one out, and at best Tolboert will avoid grounding into a double play.  

A couple things that bear thinking about though.

First, if you look at the fact that this is the average over dozens of seasons, as you acknowledge, that difference between .7 and .78 would almost have to be statistically significant given the large sample size. 

Second, if you also think about bunting as trying to come about in statistically advantageous situations, that still feeds the average. Every team is trying to do it in situations where they feel they can do better than the outcome matrix would suggest. The results still come out that way roughly because (with few exceptions) it doesn't matter what your motivation is. You will still adhere to the expected outcomes

The problem is that those statistics represent an average over dozens of years and don't tell us too much on their own. These are the averaantage they think they do.  ge runs during those situations... on average, with average hitters on average teams in average situations. We don't even know the standard devations. Are 0.78 and0.70 actually even different staistically speaking? 

The problem is analyzing bunts in their own bubble. Bunts are almost never designed to produce a run on their own; bunts are a supplementary tool to get more leverage on the bases so that future hitting events have a higher probability of scoring a run.

A bunt on its own is useless for increasing the score. But what about a bunt and a single? What about a bunt and a double? Well, you'd have to look at the "Single with a player on 2nd" or "a double with a player on second" charts to know the answer. However, how the baserunner got on second in the first place matters.

One of sabermetrics' earliest tools, Expected Runs, challenged bunting's status as a universal good.

Click here to view the article

IMHO this is a flawed premise. Expected runs is not the benchmark to use. Win probability is the benchmark to use.

What about National League with pitchers batting? They almost always hunt a runner over in those situations where a base runner can get into scoring position or to third with less than 2 outs. Is this bad baseball?

A more practical consideration in modern baseball is that not that many guys bunt well enough to make it pay. When I guy is trying to bunt but can’t get it down you’ve really screwed yourself.

Your comment is absolutely correct. But one also has to consider that sabermetrics has driven the ability to bunt, and the desire to bunt into submission. If you can hit a major league pitch, you can bunt. The fact a player  doesn’t  want to is more likely the problem. Sabermetrics cuts both ways. If you have a hitter up who statistically is likely to hit into a DP or strike out, bunting someone into scoring position makes sense. Relying on extra base hits is wonderful if you have the talents to do it. It’s like the 3 in basketball has become. Everyone knows you can score more points that way. But few seem to acknowledge that first the ball has to go into that little metal ring. For a,local example see Pitino, Richard!