Author: chandler yergin
Date: 05:53:06 02/06/06
Go up one level in this thread
On February 05, 2006 at 17:27:24, Rolf Tueschen wrote: >On February 05, 2006 at 15:44:36, chandler yergin wrote: > >>What do you think you are testing? >>Engines? >>No! >>You are Testing the current Theory of Chess. >>The Openings! There is a Direct correllation. >>Games have 3 outcomes. Win Loss Draw. >>You cannot refute over 2 million games played. >>http://www.talkchess.com/forums/1/message.html?484643 > > >I agree. To exclude the internal chess factor we had to run hundreds of games >with a single line and from both sides. I agree the more with you because I know >that the clever programmers invented the non-repetition tool so that an >automatic key disallows that a line comes more than once. In 1996 especially Ed >Schröder began that debate when he saw tests with always the same games. He >created such tools for autoplayer matches without repetitions. If he had a game >list he just deleted repeted games as if they had no value at all. I opposed >that nonsense because of course a machine that doesnt lose in always the same >line without needing such an automatic control is BETTER in strength than the >one that always falls into the same traps... I think this is evident. >But now we discuss the importance of chess lines as one of the most important >factors of a statistical result in a test. It is evident that a machine that >wins more than its opponent although the line favors a certain side or is >drawish, then this machine is STRONGER than the other. - From this trivial >reflection I argued in 1997 that this 6-game match had no validity at all >because it depended on the chance factor in chess itself. Of course already in >that year DB was strong enough to NOT fall into all too simple traps so that >Kasparov only would have a chance to win with the help of a deeper preparation. >But we all agreed that exactly that preparation he didnt have. So, basically the >result had no validity regarding strength, had no meaning at all. - > >I fear that this question is too difficult for non-statisticans so that the >meaning for all testing cant be understood by most members. Of course Bob Hyatt >always pointed out that under thousands of games one couldnt make significant >statements and in that fact the chess chance factor is the main key. > >I wished that your message would fall on fertile ground so that our test quality >could be increased. Thanks Rolf! My old friend... You are the only one (apparently)that understands and knows the Truth behind the Myths that have been perpetuated on CCC for years... cy
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.