Author: Tom Likens
Date: 07:46:20 09/22/02
Go up one level in this thread
On September 22, 2002 at 01:20:00, Uri Blass wrote: >On September 21, 2002 at 18:09:27, Severi Salminen wrote: > >>>3. winning captures ordered by MVVLVA >>>4. other moves >>>5. losing captures >>> >>>My updated move ordering is: >>>1. from hastable first >>>2. killers next >>>3. all captures, ordered by the size of the captured piece (largest first), and >>>ties broken by size of attacker only >>>4. other moves >>>5. losing captures >> >>With MVV/LVA you can never know if a capture is a losing capture. > >I guess that he first evaluates which captures are losing captures and later >sort only the good captures by MVV/LVA. > >Today I also do it(I have ideas to improve it but it is more easy to think about >ideas then to implement and it was not in the top of my priorities). > >The way that I do it today is to tell movei that if the attacker is bigger than >the victim and the square is defended by the side of the victim and not by the >side of the attacker the capture is a bad capture. > >I do not detect all bad captures by this way. > >[D]r1bqkbnr/ppp2ppp/2np4/4p2Q/4P3/5N2/PPPP1PPP/RNB1KB1R w KQkq - 0 1 > >I do not detect that Nxe5 or Qxe5 is bad because white has more than one piece >to attack e5 but at least I detect Qxh7 or Qxf7 as bad captures. > >I also may be wrong in detecting bad captures inspite of the fact that I do not >include all bad captures > >Here is a simple example: > >[D]4kbn1/6pp/5p2/4p3/4P1r1/3P1N2/4BPPP/4RK2 w - - 0 1 > >Nxe5 is detected by movei as a bad capture but it is a good capture. Hello Uri, This is a tough problem that you more than likely need the search to resolve. The capture on e5 *is* a bad capture, as you mentioned, if you only consider the value of the knight and pawn. If you want to solve this statically, it seems that you would need to check attacks that were uncovered by the piece moving, is the newly attacked piece defendable etc. By the time you do all the work for every capture, it may be easier, and faster, to keep the simple (and hence by definition, more inaccurate) test and let the search find the answer. regards, --tom > And only if >>the victim is more valuable piece than aggressor , you can be sure it is a >>winning capture - > >Even then you cannot be sure >Here is a simple example > >[D]4kbn1/6pp/5p2/4r3/4P1b1/3P1N2/4QPPP/4RK2 w - - 0 1 > >Nxe5 is a bad capture but today movei does not detect it as a bad capture. > >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.