Author: Uri Blass
Date: 12:47:33 12/27/01
Go up one level in this thread
On December 27, 2001 at 14:46:44, Severi Salminen wrote: >>The search techniques are similiar to TSCP except >>limiting the qsearch to 7 plies >> >>I believe that it is illogical to let the qsearch to explode >>and it is also illogical to let it to explode only when the >>program search deeper. >> >>Qsearch is selective search and if you search many plies >>forward then you cannot trust your score and >>if you cannot trust your score there is no logical reason >>to waste more nodes in order to find a score and >>it seems better to me to save time and to >>return static evaluation. > >In my opinion, on the other hand, it is illogical to actually limit qsearch. >1. qsearch will termainate automatically (and usually quite quickly) when there >are no captures left: so no need for depth limiting. There are practical cases when the qsearch include hundreds of nodes The following position happens in the tree of TSCP of the initial position [D]r1bqkbnr/ppp2ppp/2n5/3pp3/3PP3/2N5/PPP2PPP/R1BQKBNR w KQkq - 0 3 The qsearch includes the following captures 1.exd5 exd4 2.dxc6 dxc3 3.Qxd8+ Kxd8 4.cxb7 cxb2 5.bxa8Q?? The qsearch misses the move Bg5+ that is winning for white. My point is that when the qsearch has a long line there is a big probability that it is going to give a wrong score and if there is a big probability that it is going to give a wrong score then I do not see a reason to waste more nodes for score that is probably wrong. The point is that there are 2 cases case 1:the scores in the qsearch after more than 7 plies can change nothing about the final score of the qsearch In this case it was a waste of time to search it. case 2: correct score for the long line can change the score of the qsearch but you practically have no way to get a correct score so searching more nodes is again not productive. >to carry the position on to a quiet situation. If you limit it too early it will >surely yield wrong result by returning a static score in a non-quiet position. >It is illogical to trust qsearch the first 7 plies and then stop trusting it (if >you understand what I mean). I understand what you mean and I agree with you that trusting the qsearch completely in the first 7 plies is also not logical. I think that the qsearch should be more selective in every new ply(for example it is possible to ignore captures of pawns after few plies in the qsearch) 7 plies to stop the qsearch is not a holy number and I do not know the correct number but it is clear to me that not stopping the qsearch at some point is illogical and there is a position when Fritz needed an hour to find mate in 1 because of qsearch explosion. This position does not happen in games but I am sure that positions when Fritz wastes hundreds of nodes on the qsearch happen in games and I believe that Fritz could save time by not searching many of them. Limiting qsearch is, in my opinion, same as making >it almost useless. In most of the qsearches there are not lines of more than 7 plies so I do not see why limiting the qsearch to 7 plies make it almost useless. Uri
This page took 0.03 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.