Wired: Six Degrees of Nolan Ryan: Network Science Ranks Baseball Greats

Arguing over who’s the better player is as much a pastime as baseball itself.

Pedro Martinez or Sandy Koufax? Barry Bonds or Mickey Mantle? Of course it’s impossible to say. You can’t compare players from different eras. Heck, it’s hard enough to compare them between teams, in the same season.

But that doesn’t stop stat junkies from trying. They use equations only slightly less complex than credit derivatives formulas, and no more comprehensible to outsiders than the nose-tapping, ear-tugging, cap-pulling signals of a third base coach.

The latest entry to this field of Monte Carlo simulations and regression analyses and optimization algorithms was posted last Thursday in arXiv, an informal online repository of papers devoted to high-energy physics and self-organizing systems and other such knuckle-balling disciplines.

The study’s authors used network science to crunch the results of every single at-bat between 1954 and 2008 — and thanks to a baseball version of “Six Degrees of Kevin Bacon,” it’s possible to compare players who never faced each other.

There’s actually not much new here, from the description.  They start with “Runs Until End”, which is just probability added.  Then they take the values and do multiple recursions to adjust each matchup for the values of other matchups involving the principals.  Colin Wyers, Vinay Kumar and I were just last week talking in Primer IRC about this, and I mentioned Thorn and Palmer’s Hidden Game of Football included a wagering system proposal based on multiple recursion analysis, in 1988.

On top of not treading much new ground other than computational cycles used (of course, look at the source), the system doesn’t yet value “stolen bases, injuries...differences between ballparks” or defense.  If you’re using PA and have this kind of tech at hand, I don’t quite see the big problem including SB.  I don’t see what ‘injuries’ has much to do with anything at all.