Part Two: The bunt

In 2018, the Mason varsity baseball performed a total of four sacrifice bunts in twenty two games. I personally lead the team with two. Needless to say, it was not a significant part of Mason’s gameplan in 2018, but should it be for 2019?

The bunt has been a strategic move that modern analytics have put on the endangered species list. The goal of the sacrifice bunt is to surrender an out to advance the baserunner(s) by forcing a soft ground ball. The logic behind the bunt is, with a runner on base (normally first base) and less than two outs, to move the runner to a position where any hit allows them to score. This particularly makes sense with a lower level hitter at bat because the opportunity cost of letting them hit is low.

Simply put, modern analytics has said the sacrifice bunt is not worth it because the value of an out is too high; the value of preventing the possibility of a double play and moving a runner to scoring position (second or third base) is overrated by managers. To test this, the value of every possible bunt outcome must be calculated and their frequencies estimated. By using mean expected outcome, the overall value (in runs gained per event) of attempting a bunt can be calculated; if the value is positive, attempting a bunt would be worth it and vice versa.

I evaluated attempting a bunt to have four general possible outcomes:

1. Base hit/fielding error: The bunt is well placed and the defense cannot get any outs, advancing each runner one position.
2. Throwing error: In attempting to throw a runner out, the defense commits a throwing error resulting in all runners on base advancing two bases while the batter ends up on first base half the time and otherwise second base.
3. Failed bunt: In one way or another (strikeout, popout, etc.), no runners advance from attempting the bunt but an out is recorded.
4. Successful sacrifice bunt: The batter gets thrown out, but all runners advance one base.

Using the same run expectancy matrix from part one, I estimated the value of each event for all 24 possible configurations of outs and baserunners.

Something interesting to note: Even when a successful sacrifice bunt is performed, in only three of the twenty-four scenarios is the mean expected value positive.

To estimate the frequencies of the four outcomes, I met with Mason’s varsity baseball head coach, Adam Amerine. After explaining what each possible outcome was, I asked him to provide a rough estimate of the various frequencies as placeholders until further testing could be done (in practice). His estimates were as follows:

By multiplying each of these frequencies by the outcomes’ value, the overall mean expected value can be calculated, yielding a new matrix of the average value (no differentiation by player/batting order position) of a bunt in every scenario:

Based on the matrix, at least one of three things must be true:

1. My value estimations of the events are horribly incorrect and/or there are possible outcomes I’m not factoring in. (unlikely)
2. The frequencies of the outcomes are significantly incorrect (also unlikely).
3. The bunt is DRASTICALLY underused in high school baseball.

Of course, varying bunting abilities and parts of the batting order lead to the need for nine variations of the matrix above (essentially a 9x3x8 3D matrix). With pregame variations made to the run expectancy matrix (based on the opposition’s skill level) and the batting order, these nine 2D matrices would be on hand to serve as an aid to in-game decision making for coach Amerine.

Opposing teams could always play a more bunt resistant defense (by positioning the first/third basemen closer to home plate) affecting the probabilities of each event. However, this would make them more vulnerable to batted balls toward the first/third basemen. Either way, the knowledge of this undervalued strategy serves as a potential competitive advantage revealed purely through basic statistics.