<
>

Why head-to-head doesn't matter (much)

It was the 1999-2000 season, and I was a very unpopular guy in the state of Virginia. The beloved Cavaliers had pulled off a rare season sweep of North Carolina, yet I was insisting that UVa (19-11, 9-7 ACC, RPI No. 76, SOS No. 109) was going to miss the NCAA tournament while UNC (18-13, 9-7 ACC, RPI No. 41, SOS No. 13) was likely to make the field.

Things turned out exactly that way, of course. The Cavaliers -- and their 4-6 record down the stretch -- were relegated to the NIT, while the Tar Heels (7-5 down the stretch) found the NCAA tournament's back door as a No. 8 seed. For good measure, Carolina stayed hot and advanced all the way to the Final Four under Bill Guthridge.

This did not quell the outrage in Charlottesville, but it illustrates why head-to-head results factor only modestly into the work of the NCAA men's basketball committee. The logic -- then and now -- goes something like this:

• Individual teams are competing not just against one another for at-large selection and seeding, but against the entire pool of available teams. As such, the record of Team A against Team B is only marginally indicative of Team A's relative value in comparison to Teams B through Z.

• Even if a team is swept in a season series, those results represent only a small fraction of that team's overall performance. Think of it this way: You might get 90 of 100 questions correct on a multiple choice test, while I get only 80 questions correct. My 80 could include two of the answers you missed. The reason this doesn't give me a higher grade, of course, is that your overall performance was considerably better.

• The circumstances of the head-to-head results -- game location, point of the season, player injuries, momentum -- can and do vary widely. Results from a full conference schedule, or for that matter an entire season, generally are more accurate in determining the relative values of multiple teams.

I raise all this because, in preparing a bracket at the start of this week, Duke and North Carolina appeared as No. 1 seeds while fellow ACC member Wake Forest -- with victories over both the Blue Devils and the Tar Heels -- was relegated to the second line. My digital photo no doubt was hung in effigy in Winston-Salem.

Disregarding that Wake (at Miami) and Duke (at Clemson) have since lost, was it the correct decision to drop the Demon Deacons in favor of two leaguemates they already had defeated? I would argue "yes" for the following reasons:

• At the time of that bracket, Duke (6-1) led the ACC standings, followed by North Carolina (5-2) and then Wake Forest (4-2, tied with Clemson).

• Wake (SOS No. 80) had played the weakest overall schedule compared to Duke (No. 9) and North Carolina (No. 24). The Deacs also checked in at No. 279 in nonconference schedule strength at that point, drastically worse than the Blue Devils (No. 20) and the Tar Heels (No. 119).

• With nearly identical win totals, it was important to consider the relative damage of each team's losses. Wake at that point had fallen to Virginia Tech (RPI No. 43) and Georgia Tech (No. 146). On the other hand, Duke's losses (No. 15 Wake and No. 58 Michigan) and those of Carolina (No. 15 Wake and No. 49 Boston College) were collectively a good bit better.

It was my view Monday -- and would have been the view of this year's committee, I believe -- that the body of each team's work would slot Wake Forest behind both Duke and North Carolina. Wake's edge was in a pair of last-possession victories, while Duke and Carolina had put up more than two months of measurably superior results.

This is why head-to-head outcomes, except in cases in which the remainder of the evidence is extremely close, are mostly discounted. The race for NCAA selection and seeding is a marathon, not a sprint.

Joe Lunardi is the resident Bracketologist for ESPN, ESPN.com and ESPN Radio. Comments may be sent to Bracketology@comcast.net.