| |
BGonline.org Forums
Ughh, scratch the last sentence
Posted By: Fabrice Liardet In Response To: Ughh, scratch the last sentence (Maik Stiebler)
Date: Friday, 15 October 2010, at 12:26 p.m.
So how did you analyse the match to seven case originally? Simple, straight-forward analysis of the painfully long binomial expression for the total winning chance depending on 13 variables or is there something more elegant?
No, it was even worse than that, I computed the -2,-1 case and the way it went convinced me that it should hold for any of the induction steps. Then I thought I would appeal to the ingenious BGOnline posters to find an elegant proof, and I was not disappointed. Maybe I should have mentioned that I didn't have any rollout ?
Bob's argument is elegant, but for me the way from "every game is equally important" to "equal time allotting is correct" is still a bit foggy, although I am nearly convinced by "the more one thinks, the less the added value of an extra second".
I think I can prove it from the last assumption. That assumption means that ER is decreasing and that ER' is increasing, right ?
Then consider the possibility of allotting times t1 and t2 to two equally important games, with t1 < t2. Then ask yourself whether you should equalize the times a little by replacing t1 by t1+h and t2 by t2-h, for a very small h.
Your error rate will then be ER(t1+h) + ER(t2-h) = ER(t1) + ER(t2) + h * (ER'(t1) - ER'(t2)). Since ER' is strictly increasing, the multiplier of h is negative, so your will have decreased your error rate.
This proves that no solution with t1 different from t2 can be optimal.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.