ChessUSA.com TalkChess.com
Hosted by Your Move Chess & Games
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Move ordering
Post new topic    TalkChess.com Forum Index -> Computer Chess Club: Programming and Technical Discussions Flat
View previous topic :: View next topic  
Author Message
Dann Corbit



Joined: 08 Mar 2006
Posts: 8052
Location: Redmond, WA USA

PostPost subject: Re: Move ordering    Posted: Fri Nov 10, 2017 1:28 am Reply to topic Reply with quote

[quote="Sven"]
ZirconiumX wrote:
jwes wrote:
Deferring captures of lower-valued pieces was a loss of about 25 elo +/- 37:

Code:
Score of New vs Old: 84 - 103 - 63  [0.462] 250


Deferring captures of defended squares was a loss of about 50 elo +/- 37:

Code:
Score of New vs Old: 73 - 108 - 69  [0.430] 250


The two together is a small gain of 10 elo +/- 24:

Code:
Score of New vs Old: 230 - 213 - 138  [0.515] 581

None of these results is surprising Smile (Except that I hope that you play at least 1000 games per test run next time to increase significance ...)

Capturing lower-valued pieces includes capturing undefended pieces so deferring these is suboptimal. Capturing defended pieces in general also includes moves like PxQ where the queen is defended by a piece lower than a queen, so this is even more suboptimal. Even the combination of both is insufficient since the defender could be a queen (for instance) and you might have a second attacker of the target piece so that you could capture the queen afterwards, which does not classify the first capture as a losing capture.

ZirconiumX wrote:
[...] I haven't found the combination just yet.

There is no perfect combination but if you have not implemented SEE yet and still want to decide whether a capture most probably is a losing capture then you need a combination of criteria like this:
a) higher piece takes lower piece (viewing bishop and knight as equal), and
b) captured piece is defended, and
c) the least-valued defender has a value that is low enough to be sure that you would lose material even if you had a second attacker that could capture the least-valued defender afterwards.

Jumbo has no SEE yet. For b) and c) it does the following:
- target is defended by a pawn => losing capture
- (rook or queen captures pawn, or queen captures minor piece) and target is defended by a minor piece => losing capture

This leaves the following cases "undecided" so they are currently not classified as a losing capture:
- queen takes rook which is defended by a minor
- queen takes minor which is defended by a rook
- queen takes pawn which is defended by a rook

I should probably add support for these cases as well.

Please don't forget that you can also make use of the losing captures concept in QS: you can simply skip them there to avoid wasting time.


But be careful about positions like this, where there is a sequence of captures, especially in qs:
[d]rnbq1rk1/pp4pp/2pbp3/5pB1/2PPp3/5NP1/PPQ1PPBP/R4RK1 b - - bm exf3; c0 "give up the queen for Halloween.";
BR BN BB BQ    BR BK ::
BP BP ::    ::    BP BP
   :: BP BB BP ::    ::
::    ::    :: BP WB    
   :: WP WP BP ::    ::
::    ::    :: WN WP    
WP WP WQ :: WP WP WB WP
WR    ::    :: WR WK    


I guess that skipping the qs and reducing the depth cause problems in positions like the above.
_________________
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Back to top
View user's profile Send private message Visit poster's website
Display posts from previous:   
Subject Author Date/Time
Move ordering Matthew R. Brades Thu Nov 09, 2017 1:39 pm
      Re: Move ordering Patrik Karlsson Thu Nov 09, 2017 2:44 pm
            Re: Move ordering Matthew R. Brades Thu Nov 09, 2017 3:06 pm
                  Re: Move ordering Patrik Karlsson Thu Nov 09, 2017 4:48 pm
      Re: Move ordering Ferdinand Mosca Thu Nov 09, 2017 2:57 pm
            Re: Move ordering Matthew R. Brades Thu Nov 09, 2017 3:19 pm
      Re: Move ordering H.G.Muller Thu Nov 09, 2017 3:40 pm
      Re: Move ordering J. Wesley Cleveland Thu Nov 09, 2017 3:42 pm
            Re: Move ordering Matthew R. Brades Thu Nov 09, 2017 6:49 pm
                  Re: Move ordering Sven Schüle Fri Nov 10, 2017 12:23 am
                        Re: Move ordering Dann Corbit Fri Nov 10, 2017 1:28 am
                              Re: Move ordering Sven Schüle Sat Nov 11, 2017 11:40 am
                                    Re: Move ordering Dann Corbit Sat Nov 11, 2017 9:07 pm
                                          Re: Move ordering Sven Schüle Sun Nov 12, 2017 11:10 am
                        Re: Move ordering Matthew R. Brades Fri Nov 10, 2017 7:00 pm
      Re: Move ordering Alberto Sanjuan Thu Nov 09, 2017 3:59 pm
      Re: Move ordering Jon Dart Thu Nov 09, 2017 8:08 pm
            Re: Move ordering H.G.Muller Thu Nov 09, 2017 8:33 pm
Post new topic    TalkChess.com Forum Index -> Computer Chess Club: Programming and Technical Discussions

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum




Powered by phpBB © 2001, 2005 phpBB Group
Enhanced with Moby Threads