Author: Bas Hamstra
Date: 00:32:07 10/05/99
Go up one level in this thread
On October 04, 1999 at 14:48:41, Bert van den Akker wrote: >On October 03, 1999 at 21:48:40, Bas Hamstra wrote: > >>On October 03, 1999 at 15:08:56, Bert van den Akker wrote: >> >>>Can sombody explain what SEE is. >>> >>>Bert >> >>Static Echange Evaluation. You try to figure out without a real search what the >>outcome of a capture sequence to one particular square will be. Suppose a black >>rook is attacked by a white queen and a white knight and defended by a black >>pawn and a black bishop. Then the SEE value is determined by consequently >>letting the lowest valued piece capture to the square the black rook is on. If a >>color has performed a capture and has still gain < zero you stop. Just like you >>yourself would do to evaluate a capture sequence in a chessgame. >> >>You can determine the exact SEE value, or just if a capture is a loser, which is >>faster. Losing captures are skipped in the qsearch, which saves quite a lot of >>nodes. If you not only want to do that, but sort all captures SEE wise, you have >>to have the exact SEE score. >> >>Personally I doubt it is worth the effort of perfect SEE sorting, but skipping >>losers is a win. Everybody (except Vincent Diepeveen) agrees. >> >>Regards, >>Bas Hamstra. > > > >This concept looks very complicated because you must take in consideration pins, >capturing on different squares and check moves! > >Maybe a normal capture quiescence search is quicker and more precise. Let's compare a) doing a normal qsearch while skipping SEE losers b) and a full qsearch. The full qsearch is *slightly* more accurate. However the SEE filtered qsearch does 50% less nodes! You must consider that the SEE procedure marks captures as losers *only* based on captures to 1 square. That is something special, that the normal qsearch does not do. It is not the point SEE is faster than qsearch, or whatever. The point is a lot of captures are thrown away. Under each of which hangs a tree... Regards, Bas Hamstra.
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.