Author: Thorsten Czub
Date: 08:01:49 10/28/04
Go up one level in this thread
On October 28, 2004 at 10:26:31, Albert Silver wrote: >Check your e-mail from Ed. It's another move-ordering switch that combines well >with [MISC_03] according to him. My first tests showed problems, not just in the >finding of key moves, but in the main lines it was displaying while analyzing (I > always look closely at this), so I was unconvinced. I then decided it must be >an issue of balance and decided to make small adjustments here and there to see >if I could find it. The 109 to 108 suddenly made a difference, so I decided to >start a match with it. i gave up with most of those search switches. they more or less produce blunder when i used them, and when i knock them off, the program played normal moves. but i can try 39 of course. >Actually, I do believe it, but never expected quite such a big one. this sounds still unbelievable. >Take the Pawn Pressure switches. I've tested very high and very low with it, and >lots in between. yes. thats the way i test too. first bring the switch into the extremes. then find the right balance. but : IMO the pawn pressure switches damage the play. they push things too far. even when using 90 for the middlegame , or 110, it is IMO too much. you will see PD push the pawns like hell. also i have the feeling that middlegame is not enough STAGE. i would like to see middlegame differenciated in 3 stages. opening, early middlegame, middlegame, late middlegame, early endgame... >At more than 170 for the middlegame, the engine goes manic >about hitting pawn weaknesses. So much that it neglects other important things. >At 160 or less, it doesn't of course, but I find that it doesn't do enough. My >completely subjective opinion is that the ideal value for it is 165. However, >not for the endgame. Here I found that even values of 150 had a detrimental >effect on its play. So I've set it at 140. I mostly found this when trying to >make a special endgame setting. I can't prove anything, it is just based on my >observation. yes. we will see how it ends.
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.