Page 1 of 1
Machine learning automated GM
Posted: Fri Dec 26, 2014 11:09 pm
by Crow
New paper from Phillip Maymin
http://papers.ssrn.com/sol3/papers.cfm? ... id=2428156
Similar to work released earlier this year but not immediately clear how much is new. I dont recall seeing a consensus win metric blend before.
Re: Machine learning automated GM
Posted: Fri Dec 26, 2014 11:24 pm
by sndesai1
just skimmed through it.... did i read right that he assumed teams would have the exact same pick (draft position) available to them? or am i not understanding
Re: Machine learning automated GM
Posted: Fri Dec 26, 2014 11:41 pm
by Crow
Is it that every team would have same set of players still available for model pick as was case in real life?
First sentence of page 4 bears on this, I believe.
So the most undervalued pick would be picked by all teams above actual pick til the point where he is no longer undervalued. The perfectly modeled draft?
Re: Machine learning automated GM
Posted: Sat Dec 27, 2014 12:14 am
by Crow
May blog post is a bit more casual, easier to read on draft:
http://wagesofwins.com/2014/05/05/what- ... akes-cost/