March 13, 2007
a better rating?
Well, the Cuse isn't in the Big Dance, but that doesn't mean the OrangeJuice can't have a little fun with the brackets. I haven't decided on my picks for the Final Four just yet, though I have managed to settle on three of them at this point. I'll give more details tomorrow. As for right now, I feel an obligation to Syracuse fans to share a bit of an insider secret in the Final Four selection process.
Since the 2002 championship, the RPI has done a fairly bad job of predicting the NCAA Champion. The average RPI ranking of the Tournament Champ over the last five years has been somewhere between 7th and 8th - quite far from the predicted position of 1st. Last year, the #1 RPI ranked team at the end of the regular season was Duke, who ended up losing to LSU (RPI of 13) in the Sweet Sixteen. The winner of the tournament, Florida, was only the 15th best team in the country, according to the all-powerful Ratings Percentage Index.
So, what do we do when filling out a bracket - follow the rankings, placing statistics and strength of schedule ahead of matchups and momentum? Or, should we go with our gut instinct and pick the teams we feel have those intangible qualities that may add up to a Final Four appearance - the teams that we find endearing, with a storyline that Dick Enberg can pump up just before tipoff in Atlanta? Last year, the sentimental picks were LSU and George Mason, with a little more chalk coming in the form of Florida and UCLA - who's going to be there to take their place this time?
I believe I've found the man who can figure this out for us, or at least make it a little easier to predict. Ken Pomeroy runs his own college basketball statistics website, and seems to be making a run at Bill James status among the NCAA stat nuts - complete with a bit of contempt for those of us who prefer the traditional "system that rates teams on how 'good' their season has been" (Pomeroy suggests that we try Google or an ESPN opinion poll for that type of information).
It's to our benefit that Pomeroy has developed his own system of rating the teams, as he seems to have refined it to a point where he almost always predicts three of the teams that will reach the Final Four each year. If you've ever been in the standard office pool, you will know that having three of the four teams left at the end of the tournament will give you a significant advantage when it comes to collecting prize money.
The RPI may rank a team based on who they played and who they beat throughout the season, but Pomeroy's ratings go a bit deeper, taking into account margin of victory, strength of opposition, location of the game, and the amount of luck that affected the contest. He states that his rankings are designed to be purely predictive, assessing the likelihood of victory if that team were to play that same day.
Pomeroy's ratings have been consistently better than the RPI in predicting Tournament Success, and based on their history I can almost guarantee that at least two of his Top 4 this year will reach the Final Four (his current four are North Carolina, Kansas, Florida, and Ohio State...strangely, the top 4 seeds in the tournament, though that is uncommon. Syracuse is Pomeroy's #29). As I previously stated, the RPI seems to be a very poor predictor of Tournament Success - only once in the last five years has a team ranked in the Top 4 of the RPI won the NCAA Championship. Pomeroy, however, has had the eventual champion listed in his Top 4 in four of the last five tournaments, with the only exception being Syracuse in 2003, when he rated them #5 before the start of the tourney. I do not believe this is a simple coincidence.
For example:
In 2006 Pomeroy's top 4 teams, in order, were Florida (RPI 15!), Texas (RPI 8), UCLA (RPI 9), and UConn (RPI 3). Florida, UCLA, LSU, and George Mason made the Final Four (2 out of 4 correct), with Florida winning the championship. If it weren't for a notorious meltdown from UConn, assisting the cinderella run of George Mason, Pomeroy would've been 3 for 4.
In 2005 his top 4 teams were Illinois (2), North Carolina (6), Duke (4), and Michigan State (21!). Illinois, North Carolina, Michigan State, and Louisville made the Final Four (3 out of 4), with North Carolina winning the title.
In 2004 his top 4 teams were Duke (1), UConn (5), Oklahoma St. (6), and St. Joseph's (3). Duke, UConn, Oklahoma State, and Georgia Tech made the Final Four (3 out of 4) with UConn winning it.
In 2003 his top 4 teams were Pittsburgh (8), Kentucky (1), Arizona (2), and Kansas (6) - Syracuse (9) was Pomeroy's 5th. Kansas, Syracuse, Marquette, and Oklahoma made the Final Four (1 out of 4, though SU was 5th). SU beat Kansas in the final.
In 2002 his top 4 teams were Duke (4), Maryland (2), Kansas (1), and Oklahoma (5). Maryland, Kansas, Oklahoma, and Indiana (RPI 13 - Pomeroy had them at 7) reached the Final Four, (3 out of 4) with Maryland winning.
In every year but 2003, the National Championship went to a team that Pomeroy had as his 1st or 2nd best team in the country, despite lower RPI rankings. Not only was his system accurate enough to predict three of the programs to reach the Final Four in most years, but it could help determine which of those four would eventually finish on top. If we go with the Pomeroy Trend in 2007, either North Carolina or Kansas will win this year's NCAA Tournament - the two #1 seeds voted as Most Likely To Lose First by the ESPN.com readership. It won't be an upset, of course, but it seems most folks believe in those Florida Gators (more on them tomorrow).
I won't be using all of Pomeroy's ratings when making my picks, but it helps to have him around.
