The 27-Batter Start

I have a theory, which I have and which is mine. It goes like this: these days, when a Major League Baseball pitcher starts a baseball game, they tend to face approximately 27 opposing hitters before exiting the game. That, of course, is the number of outs in a game, and therefore the number of batters faced in a perfect game, or another game in which the pitcher "faces the minimum" over nine innings. The theory is, in other words, the idea that a starting pitcher typically has the stamina to face a number of batters that would allow for a complete game if and only if he allows exceptionally few of them to reach base. I developed this theory just from noticing, when looking over the statistics of various baseball games on MLB Gameday, that no matter whether the starter for a team in a game pitched really well, say 8 strong innings, or really poorly, allowing several runs and a bunch of hits over 4 or 5 innings, they always tended to have "batters faced" numbers right around 27.

I just did a little bit of analysis that suggests I'm sort of on to something. R.A. Dickey has made 21 starts for the Mets this year, ranging in length three complete games (for one of which I was in attendance) to a 4.1-inning blowout in the Atlanta rain in April. On average he has pitched 6.92 innings, which means he's recorded an average of 20.76 outs. The standard deviation of his "outs recorded" number for his 21 starts is 3.75, which is 18% of the average number of outs he's recorded. Now let's consider how many batters he's faced: the maximum is 33, twice, neither in a complete-game effort; the minimum is 23, also in that dreadful Atlanta start; the average is 27.33 per start; and the standard deviation is 2.85, which is 10.4% of the average. In other words, he averages approximately 27 batters faced per game, just as I suspected, and there's much less variance in how many batters he faces than in how many outs he records, just as I suspected.

Here's another way to look at it: a graph of batters faced as a function of outs recorded.

In case the text is illegible, that's the equation for the trend-line and the coefficient of determination, roughly the percentage of variance in the dependent variable explained by the independent variable. From the equation you can see that each extra out recorded on average only results in an extra half of a batter faced. From the R-squared coefficient you can see that innings pitched explains only 48% of the variance in batters faced. That's basically another way of putting my theory, that the number of batters faced in a game tends pretty strongly toward 27, in starting assignments good and bad.

Now, of course, this is only one pitcher, and perhaps an unusual one, over about two-thirds of a season. But still, I'm pleased to have a bit of empirical evidence in support of my extremely anecdotal theory.