Boyd's World-> Breadcrumbs Back to Omaha-> Go Ahead and Bunt

About the author, Boyd Nation

This week, I'll demonstrate the difference between a bad analyst and a good analyst, someone that Arthur Andersen wouldn't hire when he interviewed with them upon graduation from college (true story, although the not hiring had more to do with their lack of need at that time for programmers than anything to do with statistical analysis). I've done a study that didn't turn out at all the way I expected, and I'm going to tell you about it anyway. This tends to make for a short column when it happens, so you'll just have to bear with me this week.

One of the basic tenets of stathead thought is that the coach, during games, should sit down and shut up. Most in-game strategies, including bunting and the hit and run, are bad ideas most of the time, because the single biggest limiting factor in baseball is the number of outs you're given in a game, and it's a sin to waste them. To keep himself occupied, the coach should concentrate on his pitching staff, where he should concentrate on big-picture items like usage patterns rather than focusing on single-batter matchups, or work on his in-game patter -- things like deciding whether to challenge a kid to be a "sparkplug" or a "big horse" -- rather than bothering the hitters.

In some theoretical ways, you can actually show that bunting is a bad idea -- if you look at an expected runs chart for almost every situation, you'll find that bunting successfully improves your odds of scoring one run very slightly but largely decreases your odds of scoring more than one, and one-run innings don't tend to increase your odds of winning, especially in a high-offense environment. It turns out, though, to give away the punch line, that it doesn't seem to make any difference in actual practice.

The study that I did (those of you whose eyes glassed over in math classes should skip exactly three paragraphs at this point) was structured like this: Team OPS (on-base percentage plus slugging average) is a very good predictor for the number of runs a team will score in a season. For the teams for whom I could find numbers (255 teams, more or less randomly found) for the 2001 season, the correlation between OPS and runs scored was .909; which is very strong. The difference in the runs predicted and actually scored can be explained by a number of reasons, including random chance in the distribution of hits and the official scorer (some scorers charge more errors, which lowers the home teams OPS but leaves them with more base runners).

If bunting is truly as great a negative as I believed, then it should show up as a factor in the error, and there should be a positive correlation between how many times a team bunts and the number of runs they underperform their projected runs. Nobody publishes their attempted bunts, but the successful number of bunts is a part of the standard stat line, and I'm willing to accept that as a reasonable substitute, since it would be if the failure rate were more or less consistent. It turns out that the correlation between sacrifice hits and projection errors is almost exactly zero. In other words, it just doesn't matter that much one way or the other.

After finding this, I did find some independent confirmation of the idea in a stat called extrapolated runs, developed by Jim Furtado. XR consists of a linear predictor for runs which sums up the contribution made to scoring by most offensive events; the factor for sacrifice hits is .04, which in this context is darn close to nothing.

My best explanation for this is that, no matter how frustrating it can be for the fans, bunt opportunities just don't come up often enough to matter in the long run. There's a reasonable variation in the number of times teams bunt (I found numbers ranging from 91 down to 5 last year), but even the most of those represents less than 1.5 bunts per game. So, if it keeps the coach occupied and involved, let him bunt every once in a while just to keep his hand in without booing him (actually, don't ever boo the coach, but that's a different sermon). As always, I reserve the right to change my mind if I find better evidence.

As an aside, bunting appears to be highly peer pressure sensitive. The team that led the nation in sacrifice hits last year was Texas, coached by Augie Garrido. The second and third place teams were Long Beach State and Cal State Fullerton, coached by former Garrido assistants. For the most part, bunting seems to be clustered by conference. America East teams, on one end, bunted only 13.2 times per season, while Big West teams were all the way up at 47.0 times each on average.

By the way, Sonoma State swept the doubleheader over Hawaii-Hilo, and Jim Ward retired before last year, so somebody else is in the 20th spot on the active wins list now. Just thought you'd want to know.

Boyd's World-> Breadcrumbs Back to Omaha-> Go Ahead and Bunt

About the author, Boyd Nation