« Medtronic Wins Patent Ruling | Main | GM Sweetens Buyout Deal with New Toyotas »

Carl Bialik writes the "Numbers Guy" column for the Wall Street Journal. Today he has a fun article looking at the chances of picking perfect brackets for the NCAA Tournament.

Put it this way, it ain't gonna happen. There's even a company that provides insurance to companies that sponsor events for picking all 63 games correctly. Talk about a safe business! Fifty contests and zero winners. Papa John's contest had 90,000 entrants and no one got past the first round.

To measure the probability for this year's tournament, Jay Emerson, assistant professor of statistics at Yale, suggests using power ratings developed by Ken Pomeroy, a 32-year-old meteorologist from Cheyenne, Wyo. These ratings are based on team's records, margin of victory, strength of schedule and other factors, and are expressed in units of points. For example, through last weekend's games Villanova has a rating of 65.64 and Boston College has a rating of 61.99, so Villanova is expected to beat Boston by about four points -- the difference in their ratings -- when they play Friday.

A forecaster could use the ratings from before the tournament (which Mr. Pomeroy sent to me) to predict who would win each matchup. Mr. Pomeroy says the ratings chose a winner in about 71.3% of games this year before the tournament. "There's so much variation in performance from game to game, that even if you had a perfect system of ranking teams by how good they are, you'd still have significant errors," he told me.

Based on Mr. Pomeroy's stats, I computed the probability that teams would win in all 63 matchups -- I don't recommend you try this at home -- and found that if I had relied on power ratings, I would have had a one in 722 billion chance of a perfect bracket. (I'd also have chosen Kansas, a first-round loser, to make the Final Four.)

Of course, none of these models account for forecaster psychology. The great satisfaction of picking an upset, and the lure of picking one's own favorite team to win, combine to make picking all favorites more unpalatable than pizza is palatable. These forces conspired to make me, a writer of both a sports column and numbers column, pick first-round loser Syracuse to win the championship in our office pool. I'm tied for last place.

Posted by edelfenbein at March 23, 2006 3:19 PM