March Madness - Geek Style

I'm a geek. There's not really any debate about that; generally speaking I'd rather watch Lord of the Rings than a basketball game. I'm sure my older brothers cringe when they imagine this, but it's true. I'm a geek.

Every year I'm invited by various people to join in on the March Madness fun, and every year I can't bring myself to care enough about the teams to make any picks - often I think "maybe this year", but my lack of enthusiasm means I just don't get to it. This year however, my brother made a plea to our entire family to join in, regardless of our knowledge of the teams or rankings. I couldn't help but finally join the fun.

Of course, I couldn't just pick teams based on my familiarity with them, their jersey colors, or anything like that. No, being the geek that I am I needed a more 'Brett-like' approach. For those who are geeky like me, or those who can't help but stare at the train-wreck that is my brain (as much as you'd like to look away), I am posting the method to my Madness. Enjoy - and I really look forward to winning our family's competition and putting my more sports-oriented brothers to shame (hey, geeks are competitive too).


The Background:

I originally planned to check out the prediction markets and base my picks on what I found there, but I had no luck finding an markets that I could reference without registering... I don't want to register to a site just for this. So I decided to take a different tack. I decided to go (somewhat) random.


The Method:

Before I started to fill out my bracket, I had only two sets of non-random information I wanted to implement - (1) the seeds and (2) the historic scores for the championship games over the past ten years. I wanted to add weight for the rankings while maintaining an otherwise complete randomness (though the fact that I'm adding weight, and the fact that I am arbitrarily determining the weight the rankings should receive, really cuts down on the true randomness - but I'm going for accurate picks via pseudo-randomness, purity in randomness would provide a completely different bracket).

I decided the easiest way to incorporate the desired weight for seeds and the randomness was to use dice rolls to make my picks. So I used Random.org's virtual dice roller. This allows for up to 16 six-sided dice to be rolled at the click of a mouse, and since I needed up to 15 dice per game, it was perfect for what I wanted to do. (If you're at all like me, you'll enjoy reading the site info which talks about true randomness, the difficulty for computers to generate random results, and the site's usage of atmospheric noise to generate more reliable randomness).

To determine what a winning dice roll will look like, I decided to (1) take the difference of the seeds as the number of dice rolled and (2) require the underdog to get a '1' or '2' dice roll the majority of the time to move on in the tournament. For example, when Kansas (1) and Lehigh (16) faced off in my bracket, the difference between their seeds was 15, so Lehigh would have required a '1' or '2' on 8 of 15 dice to move into the next round. Statistically, this is difficult - which works for me because based on their seeds, it should be difficult for Lehigh to beat Kansas. Alternately, Northern Iowa (9) would only need a '1' or '2' on the single die being rolled to move on past UNLV (8) - a 33% chance.

For all inter-regional brackets (semi-finals and finals), I decided that if the teams' seeds were equal, a single die would be rolled and the winner would be chosen as follows: the teams would be alphabetized with the second in order being the 'underdog'; to win the 'underdog' would require the single die to be rolled as a 1-3, whereas a 4-6 would result in a win for the opponent - 50/50 odds here. (Unfortunately, this scenario didn't occur in my brackets, so I didn't end up using this method.)

Finally, because my brother added a tie-breaker feature, I needed to come up with the total number of points in the championship game. This is where the historic championship game scores came in; I averaged out the scores of all winners from the last ten years, averaged out the scores of all losers from the last ten years, and added those together to get my tie-breaker answer.


The Madness:

The results of my bracket were interesting, and only time will tell how successful I am. Because the results were random, I had some interesting results - I don't think any 1st ranked team in their region has ever lost in Round 1, however the dice put Vermont over Syracuse on my bracket. That's going to hurt.

On the other hand, the number of upsets increased noticeably the closer the team rankings were (desireable), and the overall number of upsets I recorded in my bracket seems fairly consistent with historical numbers. Also I found that historically it's most common for only 1 or 2 of the 1st ranked teams to make it to the Final Four - my bracket results have Duke (1) and Kansas (1), also West Virginia (2) and Pittsburgh (3). Given that it was a mere roll of the dice that got these top ranked teams to the Final Four, I'm fairly happy with my results.

After one day of games, our family standings has three people tied for 1st place, and then I'm tied for 4th place (though my Best Correct is comparatively abysmal). I've already had a few major blows to my bracket (come on, Georgetown) and there's not much chance that Vermont's going to Round 2, but overall I'm hopeful.


Final Thoughts:

I'll be back to post my final results, but so far I'm glad I took part in our family's competition. The team selection was interesting, and watching the results come in has already been fun. Depending on where I end up in the standings, I may have to use the dice again next year... or maybe come up with a different method, one that is still geeky enough.

Labels: family, sports

posted by Brett at 12:23 AM