Commentary

Last 12 games not an indicator of tourney success

Originally Published: March 14, 2008
By Andy Glockner | ESPN.com

How do we know the NCAA tournament selection committee has too much information to sort through as it shapes the bracket for this year's event? Committee chairman Tom O'Connor and his colleagues started congregating in Indianapolis on Tuesday, a full five days before Selection Sunday.

So, here's a humble suggestion for O'Connor and Co. that will slightly lessen their load: Eliminate the "last 12 games" metric from team evaluations.

On the surface, it feels like the L12 should tell you something. You'd think a hot team would be a better candidate to stay hot and play well in the NCAAs. A team that wins several high-pressure games down the stretch to force its way into the bracket would seem to be more worthy than a team that limps in.

O'Connor said he believes the L12 -- especially if its extreme one way or the other -- is worth flagging as part of a team's overall evaluation.

"My personal opinion is that it's an indicator. It makes me just pause and take a look," said O'Connor, who also is the athletic director at CAA tournament champ George Mason. "… I think it's a nice tool to have as one of the things that we use to get our attention, get the discussion going, and when you get into the last couple of teams, it would be a discussion."

But should it be? The table below breaks down the 54 at-large teams that received double-digit seeds from 2000 through 2007. These are the final teams that get in each season.

Teams Average L12 Record Average L12 Opponent RPI
Won at least 1 game 17 7.7-4.3 97
Lost 1st round 37 7.6-4.4 97

Essentially, there's no difference in the L12s of teams that won at least one game in that NCAA tournament versus teams that lost their first-round game. What if you separate BCS conference teams from the rest of the pool?

Teams Average L12 Record Average L12 Opponent RPI
BCS W 9 6.7-5.3 67
BCS L 15 6.7-5.3 70
Non-BCS W 8 8.9-3.1 131
Non-BCS L 22 8.3-3.7 115

Again, there's no statistical difference in either subgroup. The BCS teams are as identical as the overall pool while the non-BCS winners have a slightly better average record, but that record came against a modestly softer schedule. The only soft conclusion (understanding that the sample sizes are small) is that BCS conference teams tend to fair a bit better as underdogs when let into the Dance, but even then, you can't tell who's going to do what:

• There were six BCS teams that entered the NCAAs on an 8-4 run or better. Five lost their first-round game. Texas A&M in 2006 (10-2) won once.

• There were five BCS teams that won two or more NCAA tournament games. Four of them had L12s of 6-6 or worse. The only above-.500 team was Seton Hall in 2000, which went 7-5.

• There were seven non-BCS teams that entered the NCAAs very hot (10-2 or better). Five of them lost their first-round game, and only 2006 George Mason won more than one.

Given that evidence, there's significant reason to believe L12 is nonpredictive. But is selection supposed to be? Both O'Connor or Stanford athletic director Bob Bowlsby, who chaired the selection committee in 2004, said that the committee didn't lean on one evaluation tool over another, and L12 wasn't intended to be a better predictor of future performance than RPI, strength of schedule or any other criterion.

"Our task at hand is to take the cumulative assessment of what happened during the year and do a complete portfolio, because that's when the games have been played and you can't have conjecture on what's going to happen in the future," O'Connor said.

But if that's true, what is the worth of the L12? O'Connor noted that it can be used to help evaluate teams that have had personnel changes to see if recent performance foots to the overall profile. That's fair, but that period could also be six games or 20. Absent of roster change, what does finishing strong really mean if you are already evaluating the entirety of a team's profile and you're not trying to be predictive?

"There's nothing wrong with a team being on a roll at the end of the year," O'Connor said. "If that's somewhat of an indicator, it's an indicator. But I don't believe that would be the sole indicator that would put a team in or knock it out."

The committee members wouldn't need five days to do this if it were that simple. There's certainly ample reason to question whether they should be using L12 at all.

Andy Glockner is a regular contributor to ESPN.com's college basketball coverage and is the host of the ESPNU College Basketball Insider podcast. He can be reached at bubblewatch@gmail.com.