Author: Heinz van Kempen
Date: 12:57:14 10/23/05
Go up one level in this thread
On October 23, 2005 at 14:36:50, George Speight wrote: > >I would imagine the differences seen in a lot of test results depends on what >you are looking for. I think CETG does a wonderful job, with great testers. And >provides accurate results. Its' just that they are looking for different results >than I am. Starting with pre-defined positions, universal books, etc. will give >you a good comparison between 2 engines. And that is not my interest. Different >strokes for different folks, I guess. I am only interested in total program >comparison. And the only way to arrive at that is to let them use their own >books, and any and everything else they brought to the table with them. Lock, >load, play. If one program doesnt have book learning incorporated in it, that is >something the programmers will have to add or live with the results. I guess at >times either approach may very well give similar results, and at times not so. >At any rate, I dont consider either approach wrong- just a matter of preference >in what you are after. Regards, George PS: I dont doubt that I will get a >deluge of responses telling me in a nice way how stupid I am. Hi George, I agree to you, there is nothing right or wrong here. We give different pictures and all contribute to a variety of investigations, that should satisfy the authors and interested readers. Testers like you or Thomas Logan want to see how the "complete package" performs and testers like Kurt, Klaus Wlotzka and CEGT want to test the engine strength without book and bases on typical positions. There is no competition here and testers should simply stop to tell other testers that they are doing something wrong. There are simply different and valuable approaches. Our approach is to test the pure engine strength. Therefore we have to avoid book lines with 20 moves and more from own books, where often only the strength of engines in late middle game and early endgame is tested until the tablebases come into account. A proposal from Sarah is to create an own CEGT book with lines (variations) from Grandmaster practice above ELO 2600. Additionally to cover the main lines we currently have in practice this should also be variations well tested over many games and not something where still novelties change the estimation of the line being better for white or black. Like in remis.ctg this should be variations that offer same chances for both sides. Luckily we also have strong chessplayers in our team like Thomas or Uschi who could evaluate this lines and make a good choice. Best Regards Heinz http://kd.lab.nig.ac.jp/chess/cegt/ http://www.husvankempen.de/nunn/
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.