Author: Bas Hamstra
Date: 14:20:21 04/11/02
Go up one level in this thread
On April 11, 2002 at 12:55:45, Gian-Carlo Pascutto wrote: >On April 11, 2002 at 04:32:02, Bas Hamstra wrote: > >>Sounds reasonable. However I saw you complain many times about nullmove hiding >>many mate threats. Well, that IS related. With such a mini qsearch you create >>dangerous blind spots. Question is are you willing to sacrifice a little >>positional depth to get rid of those blind spots. Personally I am. > >I agree with your assesement, but whenever I tried checks in >qsearch, my search blew up. Even if I limited them in all kinds >of ways, the branching factor got out of hand. It already is >somewhat large, so I couldn't allow it to get bigger. > >I wonder if you 'try' checks (have a few rules that indicate >when it might be worth trying certain moves) or 'filter' checks >(have rules when a check is prolly bad) or do something >even different. > >Because I couldn't get it to work in qsearch, I've been trying >to make the evaluation catch those blind spots. So far, it helps >againt the biggest problems, but still... > >-- >GCP It already helps enourmously if you do it only at the first (or first 2) ply, and for the rest of the qsearch only evasions, the latter meaning you simply never evaluate when in check. Another option is allowing one and only one non-capturing check, ie a couple of checking captures occur and a non capturing check ends the sequence. Another one: do not allow more than 2 non capturing checks in a row. In all cases you must do evasions. And it doesn't work too well with futility pruning in the qsearch. Because a capture might be < Delta, but if it is a checking one with SEE > 0 it might very well give a nice result. To only skip futile non-checking captures doesn't buy you much, IMO. According to my experiments, it gets tricky if you try to do heavy things too deep in the qsearch (though I have always done evasions everywhere). Bas.
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.