Author: Robert Hyatt
Date: 09:16:55 09/04/01
Go up one level in this thread
On September 04, 2001 at 07:52:42, Uri Blass wrote: >On September 04, 2001 at 06:58:19, Adam Oellermann wrote: > >>On September 04, 2001 at 05:59:08, Uri Blass wrote: >> >>>On September 04, 2001 at 04:48:49, Adam Oellermann wrote: >>> >>>>On September 04, 2001 at 04:36:58, Jouni Uski wrote: >>>> >>>>>On September 04, 2001 at 04:13:47, Chris Taylor wrote: >>>>> >>>>>>On September 04, 2001 at 02:26:23, Jouni Uski wrote: >>>>>> >>>>>>>Bob has obviously abandoned 2 weeks version cycle, which is good!! I prefer >>>>>>>to get new version, when it's really better than old. E.g. in Junior's case >>>>>>>after 1,5 years! >>>>>>> >>>>>>>Jouni >>>>>> >>>>>>I have a feeling there will be more versions! >>>>>>If I had a large amount of money, and the betting shop was a tad nearer. I >>>>>>would, have a flutter. But then again, I am not a betting man? >>>>>> >>>>>>In auto232 matches I have 18.10 from bob's site doing rather well. It seems to >>>>>>be getteing stronger with each release. >>>>>> >>>>>I am not so sure. Look at this rating list from Herrmann: >>>>> >>>>>Crafty 17.14 2563 300 259 18 23 268.0 89.3 2207 >>>>>Crafty 18.10 2561 372 309 43 20 330.5 88.8 2212 >>>>>Crafty 17.12 2516 192 149 27 16 162.5 84.6 2225 >>>>>Crafty 18.07 2492 336 261 41 34 281.5 83.8 2211 >>>>> >>>>>Only sure thing is, that 18.7 was weak... >>>>> >>>>>Jouni >>>> >>>>A few observations... >>>>- The ratings given above are so close that the differences are probably well >>>>within the error bar. >>>Crafty 17.14 seems to be better than Crafty18.07 based on the list. >>>I do not see that the difference is in the error bar. >>> >>>>- Given also that other programmers are changing the rating pool (Blikskottel, >>>>for instance, has had probably 250 points worth of improvements in the last few >>>>months), just to stay at the same level probably indicates steady improvement >>> >>>No >>>Software changes in the same player are not allowed in >>>the list that Jouni gave,otherwise you could see >>>only one crafty in the rating list. >>> >>>If there is a software improvement >>>the new software is considered to be a new player. >>> >>>>- From what I've seen posted here, Bob seems to be concentrating on improvements >>>>in human/computer play. While this has real impact on an engine's true strength, >>>>it won't show up in the comp/comp rating lists at all, as comps can't do >>>>anticomputer style. >>> >>>I doubt it. >>>I believe that there are programs that show anti computer style. >>> >>>One of the anti computer style is to sacrifice for a king attack and ] >>>there are programs like Junior7 that has big positional score and >>>can sacrifice material for a king attack. >>> >>>Uri >> >>My reasoning here was simply that if a program had an eval which would allow it >>to play anticomputer tactics, the same eval should make it more or less >>invulnerable to anticomputer tactics. Since no computer seems to be invulnerable >>to anticomp techniques (just ask Mr Nemeth), no computer should be capable of >>genuine anticomp play. Making slightly speculative sacrifices against a castled >>king does not count as proper anticomp play, in my book. >> >>- Adam > >There are many kinds of anti computer tactics and if programs know to play one >kind of anti computer tactics then it does not mean that it is invulnerable to >another kind of anti-cmputer tactics. > >Uri The kind of anti-computer tactic I have been working on is blocked pawn positions. I haven't seen _any_ program that knows how to exploit such positions, including my own (at present). Crafty has become fairly adroit at preventing most blocked positions, but it doesn't know when to block them when it should (this will come in time). However, against computers this is worthless code since none know how to block the position, then exploit it against an opponent. Which means that simply crafty is going to spend time trying to avoid blocked positions for the wrong reasons, against another program, and end up in an inferior (but unblocked) position and lose, where it could well have drawn by being more conservative. This "change" in the eval will happen before too much longer, so that it will know when to avoid blocked positions and when to block them itself to avoid difficulties. Right now it is all of one and none of the other, which is not so great. As far as the 18.7 version goes, a couple of hundred games against computers is not enough to say whether it is better or worse than other versions. 18.7 definitely had some bugs, but then so did every other version as well. THe book, learning, and opponent's opening selection has a lot to do with how it will do in matches... probably more-so than just raw playing skill, unfortunately.
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.