Author: Uri Blass
Date: 14:59:12 11/23/01
Go up one level in this thread
On November 23, 2001 at 17:29:58, Jesper Antonsson wrote: >On November 23, 2001 at 14:31:39, Uri Blass wrote: > >>I do not see a reason to assume that intuitively there is diminishing return if >>you play the program against itself when you have not diminishing returns when >>you play with different programs. > >Well, I think there should be diminishing returns regardless of what >combinations of programs you use. I'm sorry if I haven't followed the discussion >closely, but are you claiming that you have data that says otherwise? Could you >point me to it or repeat it, please? Here are part of my results The results do not suggest deminishing returns in the nunn2 match. I have also results of the programs at depth 9 and 10 against the same or smaller depthes but I do not remember the exact results I hope chris taylor is going to use my pgn to post them. Tiger14(depth3)'s results: Deep fritz depth 3: 34.5-15.5(3 draws in the match) Deep Fritz depth 4: 18-32(4 draws in the match) Deep Fritz depth 5: 9.5-40.5(9 draws in the match) Deep Fritz depth 6: 5.5-44.5(5 draws) Deep Fritz depth 7: 2-48(4 draws) Deep Fritz depth 8: 0.5-49.5 Tiger14(depth 4)'s results: Deep Fritz depth 3: 42.5-7.5(3 draws) Deep Fritz depth 4: 35.5-14.5(7 draws) Deep Fritz depth 5: 23-27(12 draws) Deep Fritz depth 6: 11.5-38.5(7 draws) Deep Fritz depth 7: 4-46(2 draws) Deep Fritz depth 8: 1-49(no draws) Tiger14 depth5's results: Deep Fritz depth 3: 47.5-2.5(3 draws) Deep Fritz depth 4: 40.5-9.5(3 draws) Deep Fritz depth 5: 29.5-20.5(17 draws) Deep Fritz depth 6: 23.5-26.5(7 draws) Deep Fritz depth 7: 17.5-32.5(5 draws) Deep Fritz depth 8: 6.5-43.5(7 draws) Tiger14 depth6's results: Deep Fritz depth 3: 49.5-0.5 Deep Fritz depth 4: 46-4(2 draws) Deep Fritz depth 5: 41.5-8.5(5 draws) Deep Fritz depth 6: 27-23(16 draws) Deep Fritz depth 7: 19.5-30.5(5 draws) Deep Fritz depth 8: 12.5-37.5(9 draws) Tiger14 depth7's results: Deep Fritz(depth 3) 48.5-1.5(1 draw) Deep Fritz(depth 4) 46-4 (6 draws) Deep Fritz(depth 5) 43-7 (6 draws) Deep Fritz(depth 6) 37-13 (12 draws) Deep Fritz(depth 7) 28.5-21.5(9 draws) Deep Fritz(depth 8) 22-28(14 draws) Tiger14 depth8's results: Deep Fritz(depth 3) 48.5-1.5(3 draws) Deep Fritz(depth 4) 47.5-2.5(3 draws) Deep Fritz(depth 5) 46.5-3.5(3 draws) Deep Fritz(depth 6) 40.5-9.5(9 draws) Deep Fritz(depth 7) 35-15(10 draws) Deep Fritz(depth 8) 35-15(10 draws) latest results: Deep Fritz(depth 13)-Tiger14 (depth 12) 31.5-18.5 Deep Fritz(depth 13)-Tiger14(depth 13) 20.5-29.5 Uri > >>Suppose that you have a game when half of the positions A does not understand >>and half of the positions B does not understand. >> >>At small depthes tactics is going to dominate so more plies are going to help >>between A and B. >> >>At big depthes in half of the cases A is going to lose the positions that it >>does not understand and in the second half B is going to lose the positions that >>it does not understand and you are going to get 50%. > >Understanding is a function of the depth you reach and the eval. On the average, >any program understands less if you give it less depth, and there is diminishing >returns if you give it more. > >>The depth is not going to be important because at every depth that is smaller >>than 30 plies and bigger than 10 plies the program that does not understand the >>position is not going to have enough depth to solve the problem by search. > >That doesn't make sense. Every extra ply between 10 and 30 and above will make >the program gain "understanding" of some positions. > >Jesper At least theoretically the situation that I describe is possible. I agree that it is an extreme example but there is no proof that it is not possible. There are positions in chess when one program does not understand the king safety problems and depth is not important when it is not more than 20 plies(I will be careful not to talk about 30 plies because I did not try it but I am sure that there are programs that do losing mistakes even after 20 plies of search. A program that understands the position may need only 10 plies to avoid tactical mistakes and win the game. I talked about 30 plies in my example but I did not say that the game is chess. Uri
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.