|
BGonline.org Forums
How can you trust a number...
Posted By: Chuck Bower In Response To: How can you trust a number... (Bob Koca)
Date: Friday, 14 January 2011, at 3:36 p.m.
What I meant was that a short repeated trial simulation (using their engine) said P(Oregon defeats Auburn) is 3.75 s.d. above a coin flip. Equivalent to 80% wins in 25 trials. "Equivalent" because rather than simply counting wins and losses, I totaled the individual team scores and then converted the average of each of those to p(w). I think that is stronger than simply counting (more information), but I can't prove that at the moment. I just used sqrt(p'[1-p']/n) to get s.d., where p' is the converted probability of winning -- taken from the aggregate scores as mentioned above.
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.