Thursday, March 19, 2015

We're all mad here. I'm mad. You're mad.

As that most furiously-paced few weekends of the sports calendar descend upon us, there is one immutable truth about the NCAA Tournament that I am reminded of time and again:

We know absolutely nothing about college basketball.

Really. We don't. The NCAA Tournament, which tips off today, is a mind-boggling cataclysm of assumed hoops knowledge, intensive well-intention research and the massive heap of garbage that gets dumped all over it the moment Cinderella crashes your bracket. No one has any idea what will happen, beyond Kentucky probably winning, when the Big Dance starts off today, least of all me.

That, of course, doesn't stop us from trying, and to that end this year, like every year, I have filled out my bracket and made picks in my survivor pool, a pool from which I was eliminated at the end of the very first game of the tournament last year. This year I actually feel pretty good about my choices, a largely chalky Final Four with a reasonable seasoning of upsets in the first two rounds. That's generally how the tournament goes, after all, but it's the matter of which upsets you will (or more likely) won't get. This year I'm feeling reasonably confident about my faith in Buffalo and UC Irvine for no rational reasons at all, but lately I've been getting anxious as they each become trendier picks. Again, this is not in any way rational, but what, exactly, is rational about boiling a 34-game season of more than 200 teams to a 68-team single-elimination tournament?

Probably nothing.

And so, because I can't rationally think of a better way to solve and understand an irrational idea, I've decided to become one of those nerds that undertakes a totally ludicrous mathematical experiment when March rolls around, in order to determine if I'm better off using my own intuition, picking based on historical matchup probabilities, or if I'm better off flipping a coin. After going through the process, I won't lie, there's a very good chance the Coin Flip bracket, in which Arizona and Oklahoma somehow faced off in the title game despite universal even odds, ends up being the most accurate. Then again the bracket ended up with University of Alabama-Birmingham, a 14-seed that is almost certainly going to get bounced by Iowa State this afternoon, reaching the Final Four.

To perform all this experimental magic, I relied on two very helpful tools my friend Chris found through Googling after I informed him of the idea. For the bracket based on weighted historic probabilities, I used this site, which has the winning percentages for each seed vs. seed matchup since the tournament expanded to a 64-team format back in 1985. To actually perform the simulations, whether they be weighted or 50/50, I used this handy site that allows you to create a percentage weighted spinning wheel and then spin away. This, of course, isn't without its flaws. After all, as we reach the later rounds of the tournaments, certain seeds will have only played a handful of times over the past 30 tournaments, or perhaps not at all. This left me with not nearly a large enough sample size of past matchups on which to base my calculated predictions, but with nothing else to work with, I went with the past, small sample size or not. The one time I had a matchup of seeds that had never met before in the NCAA Tournament (for example, a No. 5 seed and a No. 11 seed have never, ever played one another in the current tournament setup), I just flipped a coin.

Are these trustworthy methods? Well, not really. But what else am I going to do? Just watch for fun? Please.

Here's what we wound up with.

Dave's Intuition

As you can see with this bracket, which will be my main bracket that I use in any office pools, I'm doing what everyone thinks is right and taking Kentucky to win the national championship. I don't have much faith in Nova, which loses in the Elite Eight, or Duke, which gets knocked off by Utah in the Sweet Sixteen. Wisconsin, Virginia and Gonzaga round out my Final Four, and there are a few sporadic early round upsets, but beyond Buffalo's Cinderella run, I don't see anything too crazy happening other than Albany upsetting Oklahoma. This is, of course, all based on little more than gut feeling. There are worse metrics to go by, but there are probably better ones.

Weighted Probabilities

In this case, I used our trusty adjustable spinning wheel to predict each game based on a single spin calculated using past percentages. If the higher seed matched or exceeded its historical winning percentage, it advanced. For the most part, the results seem pretty reasonable, though I'm skeptical that Maryland could upset Kentucky in the Sweet Sixteen. After all, Maryland is a team that needed a miracle comeback at home to beat Northwestern earlier this season. How could they beat a team with future NBA players on its bench, let alone its starting five? The Final Four of Wisconsin, Villanova, Gonzaga and Butler is largely believable, however, as are most of the first round upsets. Dayton reaching the Elite Eight would be quite the story, after its Sweet Sixteen run a year ago, though.

Weighted Probabilities x 100

Ok, here's where it gets a bit whacky. In terms of my weighted probabilities and coin flip brackets I did just one spin of the wheel per game. I decided to do a second bracket in which I simulated each weighted matchup 100 times, the intent being a larger sample size was likely to trend toward its historic outcomes. I assumed this would make it more accurate and we'd have a pretty chalky outcome. It soon became extremely clear that forcing favorites to match or exceed their historic probability after 100 simulations was apparently much too high of a threshold, as the bracket featured one massive upset after another. None of the 1 seeds reached the Elite Eight, with Kentucky going out to Purdue in the second round. No. 10 Indiana makes a shocking Elite Eight run, as does No. 4 Georgetown, No. 6 Providence and No. 11 Ole Miss, which, mind you, has already played a play-in game. It's here that we reached some statistical anomalies. No. 5 Arkansas and No. 11 Ole Miss face off in the West regional final, which is a seeding matchup that has never happened in the history of the 64-team bracket. In that particular case, I just flipped a coin, which probably is a terrible method since I doubt those teams are really evenly-matched, but since that 50/50 flip went Arkansas' way, I feel ok with it. Joining the Razorbacks in the Final Four were Purdue, Louisville and Iowa State in an almost incomprehensibly improbably combination. Iowa State and Purdue got through to the championship game, which presented a No. 3 seed against a No. 9 seed. This particular matchup has happened twice in Tournament history, and while I can't guarantee it's never happened in the championship game, I'd feel pretty safe saying so. Either way, No. 3 won both of those previous games, which means the Cyclones end up with a most-unlikely national title.

Coin Flips

This was the bracket in which I expected the most randomness and variability. There is plenty of that. After all, three of the 16 seeds eliminate No. 1s in the first round, a remarkable feat considering a 16 seed has never, ever won a game in the NCAA Tournament that wasn't in a play-in scenario. This makes Coastal Carolina and Lafayette's runs to Elite Eight about as amazing as they are absurd. All that said, I was almost disappointed with how not crazy things got. The Midwest and East brackets are relatively free of absurdity beyond the 16 seeds and aside from UAB's impressive run of winning four consecutive coin flips, each of the other Final Four participants is a perfectly reasonable entrant, with many predicting Oklahoma and Arizona to actually reach Indianapolis. In the end, as it turned out, the Weighted Probabilities x100 bracket was by far the zanier mix of unlikely outcomes. But hey, if UAB makes it to the Final Four somehow, well, you all heard it here first.

So which of these do I expect to perform the best over the next few weeks? Well, it's hard to say, but I think the likeliest outcome is probably my own, determined with the ruthless efficiency of gut reaction and the most minimal amount of research possible without doing none at all. And that leaves me....right where I started.

Apparently math doesn't know anything about the Tournament either. But that's ok. Neither does Barack Obama. We're all lost here. That's cool, though. After all, watching your friends palm their foreheads is half the fun of this kind of thing anyway. Now that the tough three minutes or so it takes to fill out a bracket are over, I'm ready to just sit on my couch and watch my brackets get lit on fire.

The four best days on the American sports calendar are here. Enjoy the show.

No comments:

Post a Comment