tag:blogger.com,1999:blog-3722233.post114223059198903982..comments2020-01-16T02:41:14.513-05:00Comments on Computational Complexity: March MadnessLance Fortnowhttp://www.blogger.com/profile/06752030912874378610noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-3722233.post-1142312178537806812006-03-13T23:56:00.000-05:002006-03-13T23:56:00.000-05:00The guesses simply define a probabilistic model fo...The guesses simply define a probabilistic model for the outcome. The likelihood of each model given the outcome must be your score.<BR/><BR/>Obviously I need to define a prior on the models. This would of course depend on the seeding. Unlike the usual setting where models "close" to the seeding have a higher prior, I recommend models far from the seeding should have a higher prior. This will give you more credit for predicting an upset! Of course the prior can be a tunable parameter depending on how risk-averse the office workers are. I'll leave the exact details to someone who knows a little more about this stuff.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-3722233.post-1142300665166615932006-03-13T20:44:00.000-05:002006-03-13T20:44:00.000-05:00This paper might be relevant:"March Madness is NP-...This paper might be relevant:<BR/><BR/>"March Madness is NP-hard"<BR/>David Liben-Nowell, Moses Liskov, abhi shelat, Adam Smith, and Grant Wang<BR/>http://theory.lcs.mit.edu/~cpeikert/pubs/march_madness.psDavid Molnarnoreply@blogger.comtag:blogger.com,1999:blog-3722233.post-1142236588259376652006-03-13T02:56:00.000-05:002006-03-13T02:56:00.000-05:00Minimum number of inversions between a permutation...Minimum number of inversions between a permutation that could have led to the predicted tree and a permutation that could have led to the actual tree (assuming stronger team always beats weaker one)?<BR/><BR/>Not that I have any idea offhand how to calculate that, let alone whether the calculation could be simple enough to be done by hand. But who needs to do things by hand when we have computers?D. Eppsteinhttp://11011110.livejournal.com/noreply@blogger.com