Mark Buehrle’s recent perfect game had the staff at TargetPoint wondering, is there a way we could rank those eighteen perfect games and somehow determine which was the most perfect of them all? Is there a “perfection metric” we can calculate? Here’s my attempt at an answer…
A perfect game is defined as a victory in which no opposing player reaches base (and lasts a minimum of nine innings). Under this definition, there are only two ways a player can be stopped from reaching base: he strikes out, or he is put out following contact with the ball.
This implies two paths to a more perfect game:
- a pitcher strikes out 27 batters, or
- a pitcher throws only 27 pitches followed by a put-out from the surrounding defense.
Neither of these is technically impossible, but they are thoroughly improbable. (The record for strikeouts in a nine-inning game is 20 and shared by four three pitchers, most recently Kerry Wood in 1998 in 9 innings of an 11 inning game by Randy Johnson in 2001, and twice by Roger Clemens; and the record for fewest pitches in a complete game is 58 pitches by Red Barrett in 1944).
Still, this provides two statistics that can be used in some combination to establish the degree of perfection in a perfect game. Of course, we cannot simply combine these two numbers because they are directly, linearly related: to strike someone out, you have to throw at least three pitches. Considering this, what might a revised path to a more perfect game look like?
- Throw as many strikeouts as you can, and among batters you do not strike out throw the minimum number of pitches needed to put them out.
How can this new definition of perfection guide us toward some comprehensive perfection metric? I used an adjusted pitch count, in which all strikes that ultimately contribute to a strikeout are removed. Or more simply: we removed three pitches from the total pitch count for every strikeout.
The next step was normalizing the data so adjusted pitch count and number of strikeouts were given equal weight – we used a min/max normalization. We then wanted to multiply the normalized numbers together, but first had to add one to all the scores so we never multiplied by zero. The pitchers were then ranked based upon this new composite score.
Unfortunately, there is no pitch count data for the first three perfect games in MLB history, so those of Lee Richmond, John Montgomery Ward, and Cy Young sadly remain unranked (though with only 5, 5, and 3 strikeouts respectively, they almost certainly would not have ranked in the top spots).
The most perfect game of them all according to our new metric? Sandy Koufax, with 14 K’s and an adjusted pitch count of 71 (that feels right, to have him rank tops). Koufax is followed by David Cone and the Bronze of perfection goes to Senator Jim Bunning (adding a nice political twist to all this).
Not to diminish the incredible accomplishment of a perfect game, but Mark Buehrle’s actually ends up being the least perfect of the 15 perfect games we could score. Of course, a lot of the talk surrounding Buerhle’s game was that it came against the Tampa Bay Rays, a team with exceptionally good offensive stats this year.
In that case, what happens if we add some measure of the opposing team’s abilities? After all, a perfect game is even more impressive when it comes against a good team than when it comes against a basement dweller. In a perfect game situation, in which none of the 27 batters reach base, the ideal measure of an opposing team’s strength would be their on-base percentage (OBP).
I obtained the OBP for each opposing team during the season of play, normalized to give it equal weight in the formula, and then simply multiplied the original composite score by this new measure. Even adjusting for the opposing team’s propensity to get on base, Buehrle’s perfect game moves up only one position to 14th of 15 spots.
Interestingly, adding this competitiveness measure really shakes up the top positions. Koufax, our previous number one drops to fourth place, replaced by Randy Johnson, whose perfect game came against the 2004 Atlanta Braves that finished first in the NL East and had an OBP of .343. David Cone remains in the number two spot, and Charlie Robertson, whose 1922 opponent had the highest OBP of .373, rounds out the top three.
Of course, the beauty of any new baseball statistics is debating their usefulness and figuring out how they can be improved (I’d love to see what Nate Silver has to say about this). I look forward to hearing other suggestions in the comments about how we can tweak this measure in order to figure out which perfect game is the most perfect of them all.
You can find the dataset of perfect games with pitchcounts we used here: link, and play around with it yourself. Let us know what you find!
- Alex Lundry
(baseball photo credit: http://www.flickr.com/photos/bobjudge/)