Author: Robert Hyatt
Date: 09:55:21 10/04/98
Go up one level in this thread
On October 04, 1998 at 07:41:23, Dan Homan wrote: > >On October 04, 1998 at 01:38:11, Bruce Moreland wrote: > >> >>On October 03, 1998 at 12:07:46, Dan Homan wrote: >> >>> >>>On October 02, 1998 at 17:38:47, Bruce Moreland wrote: >>> >>>> >>>>On October 02, 1998 at 16:19:08, Dan Homan wrote: >>>> >>>>>On the idea of a 'dumb as a rock' qsearch.... Mine is even dummer. >>>>>I don't use an SEE in the qsearch, but rather simply score the move >>>>>based on the value of the attacked pieces minus the value of the >>>>>attacking piece (if the attacked piece is defended). This will >>>>>avoid some captures that are sound (because sometimes we have multiple >>>>>attacks on the square), but it is quite fast and leads to a small qsearch. >>>>>I also toss out captures that lose material or don't bring me near >>>>>alpha. >>>> >>>>Basically you'll almost always check P x something, but will almost never check >>>>Q x something, right? >>> >>>Yep.... unless Q x Q or Q x something which is not defended and brings the >>>material score near alpha. >> >>OK, this isn't the pure case then. The case I just ran a test on, based upon >>your post was something like this: >> >>AxB, where A>B, never try this. >>AxB, where A=B, call it a swap for sorting purposes. >>AxB, where A >>The rest of my program was the same as it always. The difference between this >>and what I normally do is that I use a static exchange evaluator on these last >>two cases in order to get a more accurate picture of the outcome. >> >>This scored worse on the ECM tactical suite that I always run. I would report >>the exact number here but it suddenly occured to me that the machine is busy >>running another test, so I can't mess with it. >> >>bruce > >Ok, here is exactly what I do..... > >score_of_AxB = score_of_B >if (B is defended) score_of_AxB -= score_of_A > >if score of AxB is >= 0 *and* brings me near alpha then I try the >capture. I should probably use some additional logic to avoid the >'if (B is defended)' test for cases where it won't matter. > > - Dan I do this a little differently. If looking at AxB, and A is < B, then I assume gain=B-A. Else I use Swap() to calculate this as accurately as possible. The A < B simply lets me "cheat" and avoid Swap() when it is clear I am winning some material with a capture...
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.