Author: Bruce Moreland
Date: 23:35:25 11/03/00
Go up one level in this thread
On November 04, 2000 at 00:51:18, Peter Skinner wrote: >On November 03, 2000 at 23:37:35, Bruce Moreland wrote: > >>On November 03, 2000 at 15:43:50, Daniel Chancey wrote: >> >>>CMSilver is 0-2 now. CMSilver lost to CMKing+ 5.5-4.5 (+3 -4 =3) and was >>>crushed by CMQueen+(!) by a score of 7.5-2.5 (+2 -7 =1) >>> >>>I might discontinue the matches if it's clear that CMSilver isn't playing like >>>the Chessmaster Open that it won. >>> >>>The next opponent is CM6000 sel.10 >>> >>>Castle2000 >> >>This post is not an attack. >> >>Have you ever done any research in order to figure out what the odds are that >>these results could be due to chance? >> >>Here are a couple of ways to make a decision: >> >>1) Research an issue thorougly and make the decision you think is best. >>2) Flip a coin. >> >>You are trying to decide which of these personalities is best. But are you >>doing it by making a sound decision, or are you flipping a coin? And if I could >>prove to you that you were just flipping a coin, would it matter to you? >> >>bruce > >Odd Bruce that you take such a stupid stance on the matter, when just the other >day on ICC you were stating that one revision of Ferret wasn't playing as well >as another. Yet you seemed to think that everyone cared, or that everyone there >was seriously interested whether one version was better than the other. None are >public, so why should anyone really care? > >Isn't that sort of the same thing? > >We were only watching the game, just as we are reading posts on the forum here. > >Daniel is interested in ChessMaster personality settings, and their strengths, >why should he not be able to post about the tournaments he holds with them? > >If you are not interested, why read the post, and make such a lame remark about >it. Just don't read it. Don't come on and be a troll about it. > >Grow up. I'm not sure why I write a simple post and get ripped like this. Please read the first line of my previous response again. I assure you that I didn't put that in there for no reason. If you take my post literally it is not an attack on anyone. That is how I meant for it to be taken. Perhaps I did not see what was intended, but what I saw is someone doubting a version because it got smashed in a ten-game match and narrowly lost another. How often do people get this kind of feeling about versions, after matches this short? What are the odds that a version that is actually stronger will lose a match 7.5 - 2.5? These are fair questions, and if we actually discuss these questions I may attempt to find a concrete answer. Since 1998 there have been three significant versions of Ferret: 1) The version that existed from March 1998 until a week or two ago. 2) A version last week that had different pawn structure eval. 3) A version last night that had different king safety eval. If you heard me talking about Ferret having problems "lately", it may have been the version that I have been playing with and testing against for 20 months. I think that I have seen enough of that version to judge it. I don't like the way it plays. The other two versions are attempts to improve upon it. Version #2 is actually pretty good, I think. I base this not upon won-loss records but on how I see it placing pieces during games. It has a different understanding of where to put minor pieces, and for the most part I'm pleased about where it is dropping them. The most recent one is unfinished, and I may have disparaged it as such. Its king safety is based upon the idea that I award one pawn to a side if its k-side pawns are perfect, award zero if they are all absent, and award some fraction of a pawn in other cases. Anyone who has written a chess program will note the lack of open file evaluation, since I did not mention the opponent's pawns, and in fact these aren't detected. Also lacking from this evaluation is any notion of smoothing as the program progresses toward the endgame. And finally, nothing is mentioned about the location of any attacking or defending pieces (non-pawns). If I made a comment about this version, it was to explain possible weirdo behavior due to the fact that it doesn't really have a king-safety function. In fact this version has played rather well, too. I think that it is possible that my version for the past 20 months has some problems due to over-evaluation of some positional aspects, and that simply ripping all that crap out of there may have done the program good. This sounds like a fair topic for kibitzes or channel 64, doesn't it? And regarding my program not being public, it is true that my program is not available for sale or download. However, I do make it available for play to anyone who asks, and I am willing to honestly answer questions about how the program works. This is a lot more public than many commercial programs. bruce
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.