Author: Don Dailey
Date: 08:04:35 06/24/98
Go up one level in this thread
On June 24, 1998 at 00:25:59, Bruce Moreland wrote: > >I trimmed too many header thingies, the following is Don talking: > >>>I have some doubt as to whether implementing the 50 move rule in the >>>search will help in a very noticable way. It's extremely rare that the >>>program is in a 50 move situation AND a won position (which it can >>>not win) simultaneously. The 50 move rule does not even guarantee >>>the win in a won situation, it just makes a win (and a loss) somewhat >>>more likely. I have seen it happen that other programs give up material >>>to avoid a draw which causes it to lose. > >Something under 2% of my ICC games against the strongest players (IM, GM, others >rated >= 2400, which is most computers) end as 50-move draws. > >Which is not to say that the 50-move rule is important in only these games, I >have other games that get close to 50 and are drawn by insufficient material or >repetition, and I get some others where I might avoid the 50-move draw by moving >a pawn, and then the game goes onto some other conclusion. > >The most typical case, based upon no numbers but a lot of watching, is that the >program sees the 50-move case coming, but sees it too late to do anything about >it, and so the game ends up being drawn anyway. > >I don't recall a case where it gave up material to avoid the 50-move rule and >lost, but it sounds more familiar that I've seen it go the other way, where >someone else did this against me and lost. > >I'm sure that there have been cases where the program was inspired to do >something by the approach of the 50-move limit, and won the game, but I can't >recall a specific instance, and I know that I've seen a lot where I've gone >"damn" and taken the half-point. A few where I've felt that had the program >seen it coming sooner, that it could have been in position to make an effective >pawn break. > >If you have pawns, and they can do something productive, they tend to advance, >which removes the need to care about the 50-move rule. If you don't have pawns, >the odds that you can win a pawnless ending by making an opportune trade are >very low, I think. > >>>What is the general feeling on this? Does anyone feel it adds more >>>than 1 rating point to the program? If it adds more than 1 rating >>>point I might consider putting it in Cilkchess if the slowdown is >>>less than 1 percent (which would approximately cancel the improvement.) > >Good question, I don't know. > >I never bothered to ask this question, I just implemented the 50-move rule >without thinking about it. > >>>I have never played a tournament game where any of my programs were >>>ever in this situation and lost a half point for not getting desparate >>>to avoid the draw although I realize this is always a possibility. >>>I have seen it happen in test games, always on super fast levels like >>>1 or 2 ply searches. In some of these situations it was just a >>>technique problem and pawn moves or captures were not involved, it >>>just couldn't win knight and bishop vs king or some other difficult >>>ending. > >If I could give you a pill that would cause you to do one thing of my choosing, >I would give you a pill that got you on ICC or some other server. I would like >to see your program play, I would like to test it with my own program, and >others would, too. The up-side for you is that you get a lot of games of >varying degrees of quality and against various types of opponent in your email, >and you can go through them and sometimes find problems and other interesting >cases. > >bruce I'm actually considering the possibility, there a two things holding me back, one is that I can't get a machine whenever I want because our big SMP Alphas are in demand. Another problem is that I don't want it to suck my time away. I am working only part time on the chess project, I have to babysit our network the rest of the time and solve problems for people. I am first considering doing an older program of mine, but one that is pretty good and running it on a single machine. I could probably run it pretty much constantly. This will give me an idea of what is involved without going to huge trouble on a big machine I can rarely get. I could also do a single processor Cilkchess on a pentium. But this program is pretty much a dog on a Pentium. It was written for 64 bit machines and on a 64 bit machine and we never tried to make it fast on a pentium. Finally I don't know how to set this up so I would have to spend some time working through this. It looks like xboard has some support for this though. If my program runs with the xboard interface is it already ready to go on ICC? - Don
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.