ADVENTURES IN HISTORICAL RANDOMIZATION/11b: ODDS OF BUCKING THE 1-RUN HOME FIELD ADVANTAGE

Wow, what a mouthful. We haven't had a title as long as that since the last time Little Richard sang his horndog R&B classic "I Don't Know What You Got, But It's Got Me." (And it's at least an urban legend that it was listening to Little Richard on a cocktail of LSD and rum & Coke™ that prompted John Lennon to write the white-boy White Album® paranoia classic "Everybody's Got Something to Hide Except Me and My Monkey"...which, in a feat of off-kilter recursion, was covered not by Mr. Penniman hisself, but rather by Fats Domino.)

Anyway.... randomization is something we've nattered about many times, many ways, which explains our deliberately imprecise (or, mebbe, random...) numbering scheme for this particular merry-go-round with. But since sweeping up the latest crumbs from the Tango LovePie™ where we notice that the concept of "random variation" is reaching a kind of manic crescendo (or, as inveterate world traveler Linneaus Sheboygan first said, "when in Rome..."), we felt that it was time to get back in the saddle, if only to be a bed-sore for slighted eyes.
,
For reminding you how the game has changed over its history is part of understanding why certain assumptions about today's game came into being, and where all of this change is likely to take us as it continues to occur.

So, here's that quick increment of a randomizing trend that can be measured from a single component in the structure of game outcomes (i.e. wins and losses seen through the lens of home field advantage).

You may remember that it's the games decided by one run that are what lopside (that is, create most of the difference in) the home-field advantage. Historically, the home field advantage in one-run games (from 1901-2013) is .612. All the rest of the games are virtually even-steven (.505).

So it's likely that a measure of team quality that has some teeth in it could be constructed from the following event: a team winning 60% of more of its 1-run games on the road.

You would expect a team that could win such a percentage of 1-run road games would be a pretty good team, yes? At least a .500 team overall? It's syllogistic reasoning, to be sure (not quite "Asshole Logic," as Bill James so piquantly termed it while doing the rhumba with his ongoing wolfpack of jumping sharks...), but the empirical data bears it out: of the 77 teams who've done that over the last 113 baseball seasons, 80% of them had winning records; 40% had WPCTs of .600 or higher.

That's a pretty good correlation overall, but (for goodness sake...) let's not mess with the monkey. What's hidden in plain view of this stat?

When we divide these teams into those who played under the "one pennant, one World Series" rule (or what we now call "pre-divisional" play), and those who are part of the divisiveness that ensued in 1969, we're faced with some curious facts:

--The pre-division teams (1901-1968) with .600+ records in 1-run road games had winning records 92% of the time and had an aggregate WPCT of .594.

--The teams in divisional play (1969- ) matching that performance feat had winning records only 68% of the time and had an aggregate WPCT of .554.

Not the equation for "fun, fun, fun", or FIP, or better
dental hygiene... ("I'm not Marcel the Monkey and
I did NOT approve this message!!")

Put another way: pre-divisional teams meeting this criteria were losing teams only 8% of the time, while the divisional play teams have had losing records 32% of the time.

So by this one (crude, simple) measure, our lack of confidence in the predictive value in this performance signifier has increased fourfold.

That's four times as much randomization, which doubles down on the "double your pleasure, double your fun" mantra of mid-60s mindlessness that those former owners of the Cubs attempted to foist off on an unsuspecting public.

Divisional play has significantly randomized the results we see in baseball, and we would be well-served to measure it in as many ways as possible in order to provide some actual meaning for the concept of (ahem...) "random variation" beyond half-empty stat-speak.