Author: Stephen Ham
Date: 11:56:52 03/23/00
Go up one level in this thread
Dear Jouni, Chris, Uri, Paulo, Gian-Carlo and others, Thanks for the interest in the CC Challenge games. I hope you continue to enjoy these games to their conclusions. You gentlemen raised some interesting points. Here's my thoughts on them, for what it's worth. I must however make it clear that I know next to nothing about computer chess, so this may cause me to write something stupid. Please feel welcome to point out my errors...that's how I learn. Chris, you mentioned that the match has 2 objectives and listed them. However, your 2nd listed objective is news to me! The point of the match, as stated at the web site, is to determine what effect chess engines have upon Corr. Chess at the Master level and up. Can a weak player buy a super fast computer with a top notch chess engine and compete in Master level events successfully? We thought it best to first establish a benchmark for what level the machine plays at WITHOUT any human intervention. This means letting the chess engine calculate for 24 hours or so and then play the recommended line. At some later date we can try a match involving human influence over the chess engine. But for the time being, I think the most value/knowledge is obtained by seeing what level the machine alone performs at. Agreed? After all, I suspect that a very weak player, knowing his/her limits, would do exactly as we are doing...letting the machine calculate for extended time periods and then using its recommended move. Since many of us masters wonder whether we are really playing against a fellow master (human) or a machine, this experiment has merit. Uri and Gian-Carlo, this may also answer some of your questions regarding why the chess engines are only analyzing the root positions. Regarding settings, I'm not sure what settings the chess engines have. Each chess engine uses the setting recommended by ChessBase USA's President, Don Maddox. One of you gentlemen commented upon the fact that I'm being given information regarding rejected lines, evaluations, etc. That was one of my concerns too. From the beginning, I advised the operators of the chess engines, ChessBase USA's Don Maddox and ICCF's Franklin Campbell, to use their judgement to withold all data that may benefit me in the course of the game, such as secondary/tertiary lines that can transpose to the line the chess engine recommends, or whether there is a material change in its assessment of the position. Franklin has indeed withheld information in these match games, only giving me the details AFTER any "window of opportunity" has passed. Meanwhile, I like having this information only because it aids in my commentary. Giving you readers the most current data helps in your enjoyment and understanding of what's happening. Finally Chris, you raised the issue of playing style. Here's where I'd like input from you readers. My philosophy is to play the objectively "best" line. When several lines seem to be of equal value, only then do I select a line that suits me stylistically. Sometimes, this forces me to play in a "universal" style when I don't have universal skills. Still, I think that in the long run I'll improve as a player and gain better results by so doing, although in the short run I can be hurt by this attitude. Chris is correct, I've promised to play this way against the chess engines and have done so. Thus I'm playing true to form (sometimes attacking and sacing material in open positions such as Ham-Fritz 6a) rather than playing into lines that computers don't handle well (i.e. dry technical battles where my natural strengths are). I've received some friendly criticism for my approach, but, I am supposed to play as if blind to the fact that my opponents are machines (just as we are in real Corr. Chess life). I think that the final result of the match games is less meaningful than to truly test the chess engines to their maximum. Hence I've received friendly criticism for taking risks in trying to unbalance some fairly even positions in order to try to win, rather than playing safely and heading for the safe cover of endgames. Again, I'm only playing the way I always do...playing for the win and using my best judgement not to unbalance the positions so much that I actually lose. In the long run I think we all benefit by learning more about chess engines in this way. We also see more exciting chess. Finally, these games are only about 8-10 moves out of the chess engines' extensive opening books, so my preliminary thoughts are certainly premature. Still, I do have some perceptions about these two chess engines: 1)They both play at a much higher level than I expected. Still, I have minimal experience with computer chess (I have Hiarcs 3 on an old computer at home...it causes my computer to crash, so I never play against it anymore). 2) Nimzo 7.32 has a fantastic opening book...very deep in the lines played. 3) Fritz 6a assesses positions more accurately than Nimzo 7.32. Nimzo thinks it has an edge in both games when I think the opposite is true. 4) Nimzo 7.32 seems more agressive. When it doesn't know what to do (technical situations with minimal tactics), it attacks! While this can prove decisive versus humans in OTB chess, this can't be objectively correct behavior in Corr. Chess with loner time controls. 5) I think Fritz 6a is the stronger of the two chess engines. It has really impressed me in technical situations and seems to have a human understanding of what to do when no tactics are at hand. Please see my latest commentary for Fritz 6a-Ham that should post Friday evening. Either it knows that the Queenless endgame favors me or it is programmed to avoid Queen exchanges. In conclusion, of the 4 games, I think I have the edge in 3 of them while Ham-Fritz 6a is equal (although I had a nearly winning position earler, but my "human" move order allowed Fritz to find a saving defense). I value readers input, so please feel welcome to contribute. That's how I learn. Thank you, Stephen Ham
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.