Author: Dave Gomboc
Date: 19:00:28 09/15/02
Go up one level in this thread
On September 12, 2002 at 02:13:29, martin fierz wrote: >On September 11, 2002 at 23:08:19, Dave Gomboc wrote: > >>On September 11, 2002 at 16:28:37, martin fierz wrote: >> >>>On September 11, 2002 at 07:23:38, Dave Gomboc wrote: >>> >>>>On September 10, 2002 at 17:03:26, martin fierz wrote: >>>> >>>>>if you want to compare, only if you do it properly - so only if i get the >>>>>logfile afterwards :-) >>>>>and i hope you realize that even with the logfile, comparing one position is >>>>>rather meaningless. >>>> >>>>It's not my program, I can't make such a guarantee! But he is having a party on >>>>Saturday, so if I have a position I can bug him a bit to get him to log in and >>>>run it... >>>> >>>>Yes, I know that comparing one position is meaningless. >>>> >>>>Dave >>> >>>4-8 about 3 moves after 11-15 in the game is the real mistake. 4-8 wins the man >>>but loses. any normal checkers program should want to play 4-8 for a while, and >>>if it's good, it should switch away from 4-8 after some time. the position is >>>rather complicated, so you should make sure that it doesn't switch back. i've >>>changed a couple of things in cake after LV to make it find this faster, which >>>it does, but it still needs a lot of time :-( >>>just try glancing over schaeffer's shoulder to see when it switches to a better >>>move :-) >>> >>>aloha >>> martin >> >>Okay, I will try. :-) >> >>So are you saying that once you take the checker you're toast? Like a forced >>loss? Or are there drawing resources later, but it's just that taking the >>checker is quite a bad move? And what should it play instead of 4-8? >>Presumably not any move will do? >> >>Dave > >yes, there is only one move X to draw in this position AFAIK. and it's not 4-8. >i don't want to bias your answer so i'll just wait and see what you guys get :-) > >i have this book generator, i think there are also people in your group who do >this kind of stuff. i let it expand this line after las vegas, and it says 4-8 >is a dead loss, and X is a database draw. > >in the improved version, cake first wants to go 4-8 with a positive score, then >it starts getting a negative score for it, then it switches to X with a negative >score, and finally the negative score gets smaller. that's the reason i'd like >to see a search log, because from that you can see at what point it realizes >that winning the man is no good, and when it switches to X, and when it realizes >that X is probably a draw. the thing is that at first both moves look good, then >both look bad, and finally 4-8 looks real bad, and X only a bit bad. at least to >cake. so the point where a program has "solved" this position is when it >realizes that X is probably going to draw. not when it displays X as best move, >because for a long while, whether you want to play 4-8 or X is just a matter of >"taste", i.e. some parameters in the eval which have nothing to do with >understanding this position. > >aloha > martin Well, I couldn't run it... apparently the endgame database code is being rewritten at the moment. :-) Dave
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.