X

Behind the Numbers: Against the Numbers

This is the one time of the year where people take an incredible interest in college basketball statistics. Folks who don’t know their Ken Pomeroy from their Jeff Sagarin rankings are suddenly asking how valuable a low turnover percentage is and if there is any evidence it correlates with tempo despite being allegedly tempo-free. Fortunately, there are lots of smart stat people who are willing to lend an analytic hand. If that’s what you are looking for, then let me point you in the right direction. 

Sullinger & His Buckeyes Perform Well in the Metrics

I obviously place a great deal of trust in respect in Ken Pomeroy’s statistical rankings that use Pythagorean expectation-based offensive and defensive efficiencies. Well, Ken has upped the ante by running a log5 analysis of the tournament field which breaks down the expectation of a given team to reach each round. Even more fun, Neil Paine at Basketball Reference ran Monte Carlo Simulation of the tournament 10,000 times using Pomeroy’s values and posted the very interesting results. Jeff Sagarin’s list uses scoring margin and a clever use of the Elo rating system (originally designed to rank chess players) to come up with his list of things to pick. Naturally, Nate Silver can’t resist weighing in with his method of making picks, which basically does for March Madness what Five Thirty Eight did for electoral math. His system, much like his polling methodology, is a weighted aggregation of different sources like Ken Pomeroy and Sagarin’s ranking plugged in with other factors that Silver thinks are important like geography, player ranking, and pre-season ranking. The sources he pulls from are exhaustive and smart while his methodology is well-reasoned. That said, it’s worth mentioning that a dumb “wisdom of the crowds” type list, such as ESPN’s national bracket (an average of all individual brackets) tends to outperform the majority of individual brackets.

Now, here’s the question: are you trying to predict the winner of games or are you trying to win a pool? These are not the same thing and it’s important to make the distinction. The national bracket, as I mentioned, usually gets a lot of the answers right. For the big questions, common sense is usually close enough. You want to know who has the best chance of winning the NCAA? Ohio State.  Pretty much every system, rankings, and analytics have Ohio State as the best team in the country. I happen to think this as well. I also think that the four number one seeds have the best chance of making it to the Final Four. Lots of folks agree with me and lots of analytics back it up.

However, if you are trying to win a pool, don’t pick Ohio State and all four number one seeds. Making a winning bracket isn’t about being good at predicting games. It’s about being better at predicting games. Let me explain: your picks aren’t made in a vacuum, but rather in a competition. You are trying to be more successful than your peers, so there is a large incentive to make picks that will set you apart from the competition, to be willfully contrarian in some  instances. An article by Chris Wilson on Slate compares the process of picking upsets to how hedge fund managers pick assets: by looking for risky bargains with high potential pay-off. This riskier strategy actually has a solidly researched academic background and is fairly sound. The key here is not just to pick the right upsets, but also to pick the right upsets that are different from your competitors’ upset picks. In bracket making, you have to make a few risky picks to win. Now, these risky picks can certainly sink you and you might end up with the worst bracket. But in a pool of any significant size with competent competitors, conservative bracketing won’t win.

You need to be risky and contrarian. Know your opponents. If you live in Lawrence, Kansas, pick Ohio State to win it all. Are your opponents analytical stat fans? Ken Pomeroy’s website isn’t exactly a secret. Pick against San Diego State early in the tournament. The problem about all the wonderful and widely available analytics is that it’s available to everyone. It’s not hard to look at a chart and see which team experts think is better. If your opponents read up on all Bayesian methodologies and logistic regression Markov chains, you need to consider making the “dumber” pick. If two paths diverge in a wood, take the one less traveled by.

Want to Be Contrarian? Pick Princeton.

Unfortunately, this kind of thinking takes us to a weird, recursive place. What if your friends read that Slate article or this column? Your tactic of contrarianism is instead the tactic of the whole group and to adapt, you’d have to rely on some form of meta-contrarianism. But what if your opponents are now thinking of the same thing? Meta-meta-contrarianism? Ridiculous recursion on until infinity? Maybe.

There is no right way to pick a bracket. Last time I checked, none of the expert bracketologists of the the world consistently win millions of dollars in big competitions. I’m sure that typically they do very well and may often win smaller pools. Ultimately though, winning a pool comes down to a lot of luck. Oddly, I feel that the best analytic strategy is to not be afraid to capriciously embrace or reject a given system or strategy as you fit. If it does come down to luck, why not make a “crazy pick” every now and then? Embrace risk. One day someone will win by picking a 16 seed to beat 1 seed and, barring obvious gross seeding error, hardly anyone will have predicted it. Be contrary and try to pick less popular upsets. There is a good reason why everyone reads Luke Winn’s columns, but remember: everyone reads Luke Winn’s columns. Look for statistical anomalies, match-up problems, or the little hint to gain an edge if you want. Don’t be afraid to re-embrace contra-contrarianism though. After all, why shouldn’t Ohio State win it all?

KCarpenter (269 Posts)


KCarpenter:
Related Post