Author: Uri Blass
Date: 01:26:27 12/30/01
Go up one level in this thread
On December 30, 2001 at 04:18:06, Severi Salminen wrote: >>The pv is dependent on the depth qsearch and starting from the pv is not >>something very interesting. >>The history table is also dependent on the qsearch and if the qsearch is limited >>there is a bigger chance that the history table is going to be wrong. >> > >Well, I don't do those in qsearch, so that's the reason. If you do, then >limiting will of course have an effect. > >>>>0 2868842 >>>>1 3611982 >>>>2 997968 >>>>3 1564826 >>>>4 1196247 >>>>5 1214999 >>>>6 1184129 >>>>7 1153555 >>>>8 1180282 >>>>9 1153618 >>>>=10 1153584 >>> >>>I don't quite understand the numbers (they include _all_ the nodes, right?) >>>above. 0 means actually a 7 ply search with no qsearch at all? 1 means 7 plies >>>of full width and 1 ply qsearch? How can 2 ply limit have only 997968 nodes? >>>Shouldn't the number increase when increasing the limit? I'm missing now >>>something important... >> >>It seems to be a question of luck >>The program sometimes change it's mind more often or less often but my >>impression is that limiting the qsearch does not reduce the number of nodes in >>general. > >I think something might be broken. The jumps are way too big. Do you get crazy >PVs? With limited qsearch to 3 and not to 7 I can get sometimes crazy pv's. I don't understand why limiting would increase the size. And on the other >hand: if limiting doesn't even decrease the size why on earth do you limit? I limit because I thought about the worst case and I also did not finish my program and today I do not return something intelligent when I limit the pv. I may use dynamic excange evaluator that was suggested by christophe. But >I think something odd is happening or you don't count nodes right as the above >numbers are very strange? Where exactly do you increase node counters? I increase node counter after making moves(my program generates only legal moves). Uri
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.