3/18 Update: I’ve modified a few picks with the news that Ty Lawson may not be able to go tomorrow because of the injury to his toe. This news casts serious doubt about just how healthy he can get over the next three weeks, and I no longer see North Carolina as a Final Four team. I have modified my picks so that North Carolina loses to Gonzaga in the Sweet Sixteen. I project the Bulldogs to go on and beat Syracuse in the Elite Eight, which means that Gonzaga is now one of my Final Four teams. (I know, I can’t believe it either.)

This column is dedicated to the millions of Americans that will be filling out their March Madness brackets over the next few days.

You might be thinking — why should I bother listening to this joker?

Well, this is the third time that I’ve written this column and in the previous two seasons (2007, 2008), I successfully picked the winner both times.*

* Past performance is no guarantee of future results.

I’m still tweaking my method, but the crux of it is simple: Start with Jeff Sagarin’s computer rankings and go from there. Over the past two seasons, teams that had a 2+ point advantage in Sagarin’s “Predictor” category went a combined 82-15 (85%). That’s a good place to start. Even when the teams are closely seeded (within 1-3 seeds), Sagarin’s ratings are solid. Last year, in games that were closely seeded, teams with a 2+ point Sagarin advantage went 14-2 (88%). In 2007, they went 8-4 (67%). So over the last two seasons, that’s a combined 22-6 (79%). Not bad.

LOCATION

Last year, there were five games where tight (< 2 point) Sagarin matchups were won by teams with a distinct location advantage. Davidson beat Gonzaga in Raleigh, Mississipi State beat Oregon in Little Rock, Kansas State beat USC in Omaha, Stanford beat Marquette in Anaheim and Texas beat Stanford in Houston. In fact, there weren’t any tight matchups that were won by the team that was at a distinct geographical disadvantage. This year, I am going to make this my first tiebraker for tight Sagarin matchups.

SEED DIFFERENTIAL

Seed differential is also a consideration, as teams with a four- to nine-seed advantage win at about a 75% clip. The data for the previous 16 seasons was compiled by BostonSportsHub, but since they are no longer updating their site, I added the seed records for the 2008 tournament. Here is a summary of the 17 years worth of data.

So if Sagarin calculates that the teams are within two points, and there are no geographical considerations, then the next thing I look at is seed. If the differential is four or more, I am going with the better-seeded team barring some overriding factor. In 2008, this methodology was 2-1, winning the Oklahoma/St. Joseph and Purdue/Baylor matchups, while losing the USC/Kansas St. matchup. (Interestingly, all three winners had a slight advantage according to Sagarin, even #11-seed KSU.) Had I gone with KSU’s location advantage, this part of the system would have gone 2-0.

POMEROY RATINGS

Last season, I used Points Per Shot (PPS) to pick seven games and went 3-4. I still believe that PPS is a vital stat, but it doesn’t take into account turnovers, which is key when trying to determine just how good a team is. Ken Pomeroy has offensive and defensive efficiency stats that take into account pace and strength of schedule, and those are compiled to calculate his Pythagorean Winning Percentage.

Here’s how the last few winners were ranked at the end of the tournament in this statistic: Kansas (1), Florida (2), Florida (1), North Carolina (1) and Connecticut (2). Clearly, when picking the overall winner, we don’t want to stray too far from this ranking.

Let’s take a look at the Final Four participants for the last five years and see how they finished, keeping in mind that their final ranking does take into account how they performed during the tournament.

2008: Kansas (1), Memphis (2), UCLA (3), North Carolina (4)
2007: Florida (2), Ohio St. (4), Georgetown (5), UCLA (6)
2006: Florida (1), UCLA (3), LSU (10), George Mason (23)
2005: North Carolina (1), Illinois (2), Louisville (5), Michigan State (7)
2004: UConn (2), Georgia Tech (7), Duke (1), Oklahoma St. (3)

So, excluding the outlier (George Mason), the average Pythagorean ranking for Final Four teams over the last five years has been 3.6. I wish the site showed the pre-tourney rankings, because it would be helpful to know where these teams were ranked when they started the tournament. Since all we have to go by is where they stand now, it would seem unwise to pick a team outside of the top 10 to reach the Final Four.

I used the Pythagorean method back in 2007, and through the second round of the tournament, it had picked 37 of 48 winners. I stopped using it at that point, and I’m not sure why. This year, I’ll keep track of its accuracy throughout the end of the tourney, though I think it’s important to use the static, pre-tourney rankings because that’s all we have to go by when we fill out our bracket.

We’ll see how much I use this statistic as we dig into the bracket.

So, without further ado…

Read the rest of this entry »