« Bobby Brown Goes Left | Main | In My Defense »

March 15, 2006

A Tax On People Good At Math

NP: The Beatles, 1

Every year, when the NCAA Tournament rolls around, I get this notion that I'm going to build a statistical model of some sort that will pick all the games for me. I think I've tried it before, but I've never been entirely satisfied with the results. I think I was still in it up until the Final Four last year, but there were enough misses to dismiss the process.

This time around, I'm feeling a little better, mostly because I found the NCAA statistical motherload, which included the top 50 teams by rebound differential. As a former center and power forward, I've been arguing for years that the team that wins the glass wins the game, over nearly everything else. So being able to factor that into the equation gives me a bit more hope than usual.

Anyway, it's not the most complicated model in the world, and it's driven pretty hard by what I think is important in the game. Winning percentage multiplied by free-throw attempts per game multiplied by rebound differential, divided by opponent's field goal percentage. I originally didn't have winning percentage in there, but having to substitute average values for teams for which I couldn't find any data caused a bunch of the scores to bunch up, and I think that statistic is more important than points per game or point differential, because neither of those really matter if you don't win the game.

So, what did I get from all this? Final Four will be Texas, Gonzaga, Connecticut and Oklahoma, with UConn beating Texas to win it. If I'm wrong, I've wasted more money on sillier things, so no real harm done.

And seeing as how I've been neck-deep in predictive modeling at work lately, I don't even see this as the lost productivity that is going to be reported ad nauseum this week.

Comments

Let's take a stroll back to our undergraduate pysch years, shall we? Remember Clark Hull, he the behaviorist/learning theorist who tried to come up with a mathematical equation that would explain the Stimulus -> Response pathway in humans? He started with just a couple of variables, such as O that stood for something like "all variables within the organism." When that didn't yield reliable or valid results, he kept adding more and more variables to the equation, such as "habit strength" and "drive", and when these didn't work he added subvariables to all, complete with supercripts and subscripts.

God I loved that man.

Well, I of course tried to do a similar thing as you and Hull, but trying to predict baseball games instead of basketball. I can't remember all the variables now, but they included such obvious things such as home winning percentage, road winning percentage, ratio of errors committed to errors by opposing team, ERA, etc. Common, obvious variables. When this yielded results that were less than perfect, I of course added some other variables. How about game temperature? Then, to get better results still more variables were added... with the thought being that if only I could add enough variables, and if only I could weight them appropriately, I'd be able to predict who'd win the division easily.

Of course, even with an infinite number of variables this doesn't work, because one forgets that sporting events, like the humans between the Stimulus -> Response link, are infinitely unpredictable.

To put it more clearly, as much as I'd like Texas to go to the final four, *my* model (i.e. "simple hunch") says that they'll lose in the reginal semis to Iowa, "rebound differential" be damned.

As Skinner would say, put that in your black box and smoke it... ;-)

There's also this, not to rub salt in a wound or anything:

http://tinyurl.com/zsdp4

Post a comment
Name:


Email Address:


URL:


Comments:


Remember info?



about notabbott.com

what is it?

notabbott.com is not spamming you -- please read

however, if you'd like e-mails about upcoming shows and whatnot, click here

recent entries in MAIN

Domino Effects
March 4, 2015

Housekeeping note
January 2, 2014

Slacker Profiteering
July 7, 2013

In My Defense
June 20, 2013

When A Foul Isn't A Foul
February 5, 2013

archives by month

credits

Creative Commons License
All content on this website (including text, photographs, audio files, and any other original works), unless otherwise noted, is licensed under a Creative Commons License.