View Full Version : Pomeroy on the ACC tournament

03-07-2007, 09:12 AM

Interestingly, three of the four teams most likely to win have to play Thursday: Duke, Maryland, and Georgia Tech. Pomeroy gives us a 17.9% chance of winning it all, but notes that the statistics do not take into account the loss of Henderson on Thursday.

03-07-2007, 05:06 PM
Of course, those probabilities are generated via his offensive/defensive efficiency ratings from the entire seasons' work. This includes in Duke's case games against Georgia Southern and Kent State, although the ratings are adjusted based on the strength of schedule.

If, however, the probabilities are re-run using only statistics from in-conference play, UNC becomes even more favored, at 86.5% EDIT: This was incorrect. UNC is a 57% favorite. Duke is still second but drops to 8.9%, followed by UVA, Maryland, and Virginia Tech (all around 8% chance). While it is still true that some teams (like Duke) played harder schedules than other teams (like UVA), I don't feel that adjusting the efficiencies for schedule strength would make much of a difference. The bottom line is that Duke's in-conference performance was decreased compared to the entire season (as represented by Pomeroy's predictions). It also reiterates the parity of the league (after UNC, that is).

03-07-2007, 05:10 PM
I realize that Pomeroy's numbers are based upon the results of his efficiency ratings, but I still find it hard to believe that a team which only went 11-5 during the conference schedule is such a heavy favorite to win it all. I haven't really looked into the statistics of it all, and maybe all that this large margin means is that no other team is actually expected to win so UNC is a heavy favorite by default.

03-07-2007, 05:17 PM
In four of their five losses (two against VPI, NC State, and Maryland), were by a combined 13 points. Thus, they aren't too far off from being 15-1, which would make them a clear favorite to run away.

The thing that they have going most against them is that a matchup with the Terps in the semis.

03-07-2007, 05:42 PM
I made a mistake in reporting the conference-only log5 predictions. UNC has a 56.6% chance of winning the tournament, thanks to a very high Pythagorean Winning Percentage. I mistakenly gave the chances for them to make the semifinals. Sorry for the confusion.

I posted the entire table on my blog for whoever wants to check it out. I am in debt to Paul Rugani for his ACC only efficiency stats. He says on his blog (http://dbdhoops.blogspot.com) that he is going to make some kind of schedule adjustment and re-run the predictions. I'm looking forward to those.

HK Dukie
03-07-2007, 08:26 PM
these statistics don't consider the extra fatigue that first-day teams must endure in the following rounds....thus the stats would tend to overstate the winning percentages of seed 5-12 and underestimate seeds 1-4 by a slight margin