Last week, during the halftime show of one of the Big Monday games, the ESPN talking heads were discussing certain teams' post-season prospects.  I can't remember the exact team they were debating about, but during the conversation Bilas pulled out a line similar to "there is absolutely no correlation between how a team performed in their last 10 games and how they perform in the tournament".

Now granted, I've become extremely skeptical of anything that Bilas says lately, especially after his tirade about Michigan that was clearly motivated by his loyalty towards former Michigan coach (and former teammate of his at Duke) Tommy Amaker.  But even if I didn't feel this way, I think I'd have looked at this statement with the same skepticism.  Could this really be?  How could he say such a thing?  Everyone knows that it's terribly important to finish strong before the tournament.  Could he possibly be right?

Well, I decided that I had a free hour or two this week, and I had just enough curiosity to look into this really deeply.  So I pulled together data that was readily available for teams that qualified for the NCAA tournament over the last nine years to see if he was right.  The end result: he's pretty much wrong.

(quick stats lesson: I'll refer to the term correlation coefficient.  All that measures is how closely two items are related.  The closer that the coefficient moves to 1, the closer the two items are to being directly related to each other.  A coefficient of 0 means that the two items have no direct relationship.)

Here's the file with the raw data I used if you want to play along as well:  ncaa tourament history 1999-2007.xls (1.13 mb)

So first I calculated the correlation coefficient of the number of wins a team had in its last 10 games to the number of wins they had in the tournament, which took all of about three seconds.  The result: a coefficient of 0.0824.  My lord, Jay Bilas is right!!  The two are almost completely unrelated!  Welll...

In the words of another ESPN commentator, not so fast, my friend.  You see, this calculation includes all teams that qualified for the tournament, including the bottom seeds.  There were six teams from 1999-2007 that won at least nine of their last ten games.  But do you really think that this helped them go up against the Duke's or UCLA's of the world?  So, to have any meaningful analysis of this, you really have to drop the clear bottom teams out of the equation.

Looking at just the top 12 seeds, the coefficient rises to 0.2316.  Closer, but it still doesn't tell us a lot.  But, if you cut it down to just the top 10 seeds - who win 88% of their first round games and have over 90% of all tournament wins over this time span - it rises to 0.3174, and at that point we can say that there is at least some relationship between how a team performs over its last ten games and how it performs during the tournament.  In fact, the only aspects I could find that had a closer correlation to tournament success were wins over top 50 RPI teams, and road/neutral court winning percentage (both of which are also looked at closely by the committee).

Now, I'll concede that it's pretty hard to isolate any one variable and say that it definitely relates to tournament success.  There's a real chicken-or-the-egg aspect to this (i.e. is a team good because they won ten games against the RPI top 50, or did they get those ten wins because they're a good team).  But, I think we can say that it's reasonable to say that how a team finishes the regular season/league tournament has at least some impact on how far they go come tourney time.

So, Jay Bilas: incorrect.  And that makes me happy to say it.