Pattern Recognition

Scott Adams has a good blog post on persuasion vs. “science” when it comes to “climate science” (known around these parts as Global Weather).  Adams makes the simple point that it’s pattern recognition, not science as such, that convinces most of us.  The reason being, most of us don’t really get the science.  We can’t do the math, so we have to rely on the track record of those few who can…. and their record stinks.  Adams proposes a three-part test to change his bias (his word) re: climate “science.”  The first, especially, is gold:

Convince me that complex models such as the climate science models have done good jobs in other fields in the past. And the examples have to involve human judgement in the inputs, and lots of iterations. And those models have to have succeeded in predicting the future five years out, or better. If such things exist in other fields, I can be persuaded that climate scientists can do it too. (No fair picking physics models. Those are not filled with human assumptions.)

I know way less math than Scott Adams (he’s got an MBA and did financial modeling for years), but even I see the problem here.  It’s that “human judgment” part.  Humans are endlessly adaptive.  That’s why we’re so evolutionarily successful.

Baseball is a fun example.  Back in 2003, a neat little book called Moneyball hit the shelves.  It’s a 288-page primer on our ability to miss the forest for the trees.  Throw in the hilariously overwrought reactions to the book (and subsequent film) among baseball people, and you can pretty much throw every basic psych textbook out the window.  We can’t go too deep into it here (though that would be a fun series of posts if y’all are interested); the point is, Moneyball’s focus on the quantitative aspects of baseball performance led to a brief fad for crazy, hitter-specific defensive shifts.

ShiftsThat’s a shift put on for Adam Dunn, a famously free-swinging left-handed power hitter.  Every season, Dunn was routinely found at the top of the leader board for home runs, walks, and strikeouts.  There’s math to back it up, of course, but just knowing those three things suggests that if Dunn manages to make contact, it’s going hard to the right side.  If he drops one down the third base line, it’s an accident… and a happy accident at that, since normally when he catches one it goes over the right field fence.  So why not stack the right side of the infield, to guarantee that anything he puts on the ground gets caught?

You don’t have to know baseball to know what happened next… though you do need to understand basic human motivations, so drive-by Lefty readers will want to pay attention here: Dunn adjusted his approach at the plate, and pretty soon nobody shifted for him anymore.  The guy’s a professional athlete, see, which means he doesn’t have a job if he doesn’t get on base.  So for a while there big, free-swinging, homer-blasting Adam Dunn turned into a dinky-dunk singles hitter… until he forced the defense to switch back to playing him straight up, at which point he went back to jacking longballs again.

Pattern recognition is dynamic.  The other guys are looking for patterns, too, and when you spot one of theirs, pretty soon they’ll spot your adjustment…. so they adjust, and you adjust, etc. etc.  The first guy to spot a pattern enjoys a brief run of great success, but it never lasts.  It can’t.  Which is why they’re still playing baseball, despite the “sabermetrics” guys’ best efforts at making a day at the ballpark as much fun as doing your taxes.

And that’s baseball, a game with millions of dollars on the line for every single player, coach, scout, and manager, each and every season.  If they can’t come up with a surefire repeating system, it can’t be done… and it surely can’t be done by “climate scientists,” all of whom have tenure and government grants and get paid — and, of course, lionized by Lefties who fucking love science — no matter if they fail to predict six o’clock at five thirty.  Pattern recognition, see?

Leave a Reply

Your email address will not be published. Required fields are marked *