#1s in Close Games, or Is Florida Fine?

Everyone has Florida going far.  Half of the entrants of Yahoo's bracket challenge had the Gators coming out of their region to the Final Four, and over 36% had them winning the National Championship.  Even Ken Pomeroy's computer guaranteed us that they were the second most likely champion in the field.  And why not?  The team is deep, talented, experienced, and good on both ends of the floor.   Plus they've won 26 straight heading into the tournament.

Then you watched them play today against Albany, the fourth place team out of the America East conference, one which had to win Tuesday night in Dayton just to get to Orlando to be the ritual sacrifice to Florida.  Only Florida didn't quite manhandle the Great Danes the way everyone expected.  The game was tied several minutes into the second half, and though Florida pulled away late, they didn't cover the Vegas line of 21.5, or even kenpom's more modest 16 point MOV prediction.

Same thing with Virginia.  Coastal Carolina led them at the half, and while UVA played much better after the break, they didn't take control until the latter half of the period when they pulled away from the Chanticleers.  It would be hard to think that the average viewer walked away from that game impressed by Virginia's prospects to beat teams which are actually supposed to be competitive with them later in the tournament.

So if given the chance, would it be a good move to scramble and scratch out those predictions of Florida or Virginia in the later rounds of the tournament because of their lackluster performance in their first game?



Well, maybe not.  There's no real clear relationship between how well a #1 seed does in Rd 1 and how many wins they'll eventually have.  R-squared here is just 0.05 so little relationship in the data from 1985-2013.  Kansas beat Prairie View A&M by a staggering 58 back in 1998 before losing two days later to Rhode Island.  On the flip side, Illinois' stacked team led Fairleigh Dickinson by a point at the break and won by just 12 before playing for the National Championship just over two weeks later.

Of course looking at the entire sample of games may obscure some of what we're trying to find out.  The difference between a 20 point win over an under-seeded team might be no more predictive than a 35 point win over a sub-.500 team which got hot in its conference tournament.  So what I did was to look at only the truly close games.  I defined "close" as a) MOV of less than 15 points and b) a deficit of 7 or less for the underdog at halftime, i.e. competitive in the first half.  I am sure that this definition may miss a case or two which could be called close, and is subjective at the very least, but on we go.

On the average, a #1 seed will win 3.37 games in the tournament, so another 2.37 after they win their 1-16 game.  The 17 cases that fit the above criteria averaged about half of a win less, at 2.82 wins.  Of note was that the group of 17 included four of the 15 #1 seeds which lost their second game, as compared to 11 of the other 99 one seeds losing their next game after playing a "noncompetitive" first round game.  The sample size is hardly large, but the difference is one in four compared to one in nine.



So the verdict?  It's tough to draw too many conclusions, but we'd be hard pressed to say it's a good thing to let a 16 seed hang to close