Author: Robert Hyatt
Date: 11:49:44 12/29/97
Go up one level in this thread
On December 29, 1997 at 14:40:40, Don Dailey wrote: >On December 29, 1997 at 13:28:29, Chris Whittington wrote: > >> >>On December 29, 1997 at 13:23:25, Rajen Gupta wrote: >> >>>when I learnt chess i was told that opening books consist of a database >>>of moves that are played at the initial phases of a game.These moves are >>>supposed to have been developed over the years in strong GM or IM games >>>and consequently have been (or at least should have been)subject to >>>extensive analysis.As a result of this, those lines which are inferior >>>for one or the other side get discarded whilst those lines that lead to >>>an approximately equivalent position get accepted into general chess >>>theory.The opening books provided with chess programmes should therefore >>>result in an equivalent position on the board after which it is upto the >>>chess engines to show there stuff. >>> >>>MCP would in fact be doing a great service to chess fans if it could >>>identify the weak lines in existing and established chess theory and >>>provide alternate moves or lines earlier on(which are superior to the >>>existing lines) rather than merely displaying a large volume of cooked >>>lines which does not in any way enhance the prestige of the programme. >>>If I use such a programme I would never know whether it is providing the >>>best possible analysis for my money or whether I have got something >>>which seems to be strong only because it knew the questions the day >>>before the exams! >>> >>>Rajen Gupta >> >>It looks like a good anti-PR demolition job has been done on Mchess. My >>problem is thta I'm not sure this is entirely fair. Its partly due to no >>defense appearing to the attacks, maybe. >> >>Chris Whittington > >Hi Chris, > >I have no problems with Mchess. I'm considering both sides of this >issue of cooked books and trying to do this in a detached way. I'm >interested in the phenomenon and not attaching blame on anyone. > >As I've mentioned before, although I'm not sure it's "ethical" I'm >not saying for sure it's not either and it's hard to fault those >who MIGHT be doing it. > >If I had a "hot" program on the list and felt others were doing this, >I might very well feel forced to defend myself with cooks of my >own. And this happens of course in master chess too. The real >problem isn't the cooks, it's the inability of programs to defend >against them on their own. > >But I am rethinking this issue all the time. I am trying to remain >open minded and do not think this thing is so clear cut. I'm not >defending it but I'm not condeming it either. > >It could even turn out to be a good thing. The rules are simple, >anything goes and now programmers will be forced to defend against >this practice which might very well turn out to be a good thing >for all of us. Maybe we should consider this a fact of life and >deal with it? It's not like there are NO defenses to this >practice. It might turn out that doing this stuff makes your >program vulnerable too. If we simply consider this another part >of a complex equation then there is no problem. Perhaps it will >become well known this this program has a great book, mediocre play >and this other program has great play, bad book etc. It should >be considered another stylistic aspect of a programs personality. > >On the other hand I like the idea of having a special rating list >for programs with completely randomized and very shallow openings. >This would give us an idea of a programs overall playing ability >on its own, and give chess owners more data to think and argue >over! > >-- Don the only issue I see is # of games. IE anyone can cook a book line for Crafty, and get away with it exactly *one* time. But then, thanks to book learning, it won't play that line again. So to bust Crafty badly, someone is going to have to cook hundreds of lines, which is *very* time consuming and difficult to do. I'd say that decent book learning can take the edge off of this problem entirely. As it should... and it will cure the problem on its own, without needing lots of hand-tuning. That's why I worked on the learning issue for a while with Crafty, because I was tired of having to hand-tune our opening book each year for the annual computer events, to avoid getting "cooked". It isn't a problem for me, even knowing that if you find 10 cooks you can win 10 games. Unless we are talking about a small set of games. On the SSDF, this might be a problem at first, since learning is not going to be "shared" even though Crafty already has the facility to share learned opening lines with other crafty copies. However, I'd say this is a non-issue for those that learn, a big issue for those that don't, but one that is repairable with some work...
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.