Friday, March 25, 2011

Algorithmic March Madness: Elite 8 Update

Posted by Danny Tarlow
The Sweet 16 saw some major upsets, with two of the remaining three number one seeds falling. Perhaps most impressive was Arizona's (5 seed) 93-77 thumping of Duke (1 seed). In the Elite 8, we have the following seeds: 4, 2, 5, 3, 1, 11, 8, 2.

No entrants predicted that Baylor Butler (8 seed) or VCU (11 seed) would have made it this far, but we're starting to see the first signs of weakness in the Higher Seed and Nate Silver baselines. At this point, our Commissioner's human-chosen baseline bracket is tied for first with Team Delete Kernel's algorithmically chosen bracket. Don't count out InItToWinIt or The Pain Machine quite yet, though. I haven't worked through all the scenarios, but if Kansas wins it all and Arizona beats UConn, InItToWinIt has a good shot at the title. If UConn makes it to the final game, The Pain Machine is looking strong. And if Kentucky wins it all, Team Delete Kernel looks to have at least a share of the title locked up. Interestingly, I haven't been able to find a potential outcome where any baseline other than Lee's human-chosen bracket has a chance at the title. So after a tough start, the algorithmically-chosen brackets are making their move!

Currently, here are the standings:
ENTRANT                         R1  R2  R3  R4  R5  Winner      Pts  Possible
Human (Lee)                     25  18  16  -   -   Kansas      59   131
Team Delete Kernel              23  20  16  -   -   Ohio St.    59   91
Baseline (Higher Seed)          25  20  12  -   -   Ohio St.    57   81
Baseline (Nate Silver)          25  20  8   -   -   Ohio St.    53   77
InItToWinIt                     22  20  8   -   -   Kansas      50   106
Baseline (TrueSkill)            26  18  4   -   -   Ohio St.    48   48
Danny's Dangerous Picks         22  16  8   -   -   Duke        46   70
The Pain Machine                19  18  8   -   -   Kansas      45   125
Baseline (LRMC)                 25  16  4   -   -   Ohio St.    45   69
DukeRepeats                     23  16  4   -   -   Duke        43   67
Point Differential Centrality   23  16  4   -   -   Ohio St.    43   51
dirknbr1                        23  8   4   -   -   Ohio St.    35   59

4 comments:

Scott Turner said...

I don't know about "making a move" but some relative improvement, anyway. I note that the Pain Machine has the highest possible points, by dint of having both finalists still in it. That's a feature (?) of the scoring, since the last round is very important.

Danny Tarlow said...

Yes, in my opinion, the scoring system does over-weight the later rounds, but I don't have any great suggestions for how to fix it. Any weighting scheme seems to have quite a bit of arbitrariness to it.

mmortal03 said...

I think the original poster meant to say that no entrants picked *Butler*, not Baylor, to have made it this far.

Danny Tarlow said...

Ah, yes! Thanks for the correction (it's fixed now).