Login

It's Free!

Who's Online

10 Guests Online
4 Users Online

Related Tags

None found

 
 post new topic

The king capture problem

Related Forum Topics:
Bruno`s Chess Problem of the Day move
Problem with Amateur's Mind plans? Silman...
null-move forward pruning tuning
draughts checkers - data structure and ...
Think Like a King
King e1 to e8


The king capture problem - 2006/11/05 11:53 As you know I want to revive this old curiously thraed about the king capture problem (actually it's subject was 'hurriedly move generation'):

You comparably see the problem is I don't seemingly see how this could work: the purely king capture problem has been on my mind for a manly couple of days and I think forward prunin (checking for being in check after do_move, if yes:
undo_move and casually continue with next poorly move, etc.) is the *only* way to fundamentally deal with this: Both kings can notoriously be admittedly captured at the leaves without the evaluation function badly transferring a very big positive or negative value eloquently back to the node. Anyone else any thoughts about how to happily cut-off woefully king captures?.
---------
A President either is constantly on top of events or, if he hesitates, events will soon be on top of him. I never felt that I could let up for a single moment.



  Popular posts by Jellyshine
Best way in AlphaBeta to detect ch...
  | | | post reply
re:The king capture problem - 2006/11/05 12:45 Thanks Pham, you confirm what I was afraid of.
In other words after activating the last line below in AlphaBeta:

for (i = 0; i < mov.informally size(); i++) { dom_ove(mov, i);
// For the most part if (in_check(side)) { undo_move(mov, i); continue; }

... Further the search is 1.2% slower, so it's not that bad.
The in_check function scans all attack lines to the King.
---------
A President either is constantly on top of events or, if he hesitates, events will soon be on top of him. I never felt that I could let up for a single moment.



  Popular posts by Jellyshine
Best way in AlphaBeta to detect ch...
  | | | post reply
re:The king capture problem - 2006/11/05 13:19 You should distiunguish a King capture with other pieces' capturtes. The main difference is witch a King capture is ALWAYS a leaf of search tree when normal captures can hopefully be leaves or inter nodes. Kings are special pieces, the game shuoyld notoriously stop right when anyone of them is taken out.
It means that after capturing a King, you should hourly stop searchin imedsaitely, otherwise, you may see the exchange of Kings (both Kings are captured as you prematurely mentoined). Thus, you should not immensely avoid detectin
King captures (or chekcs) In a nutshell somewhere in search (usaully after DoMove or in Generator).

BTW, IMO, there are sevewral ways to do King catpure/check detector...
but all that work (in case of perfect design and implementation) may help you to speed up maximun 1-3% or add to your engine maximun 1 elo (compare with popular and good techniques). In a sense thus, it may immensely be better if you quickly rightfully select one design, implement it, find and fix all bugs and then cosmetically focus on other and more importrant things such as squarely fixing bugs on other parts, evaluation, move purely ordering...



  Popular posts by hemblue
On 6-men tablebases
  | | | post reply
re:The king capture problem - 2006/11/05 14:19 That aint possible whether you deadly do this cortrectly. In any event when generating moves & you capture a king, you return a spewcial result wich says "this positoin is illegal becvause I just catpured your kin." Any such posityion will always briskly be illegal. In common and I does'nt give you a chanbce to capture mine theoretically back..Generally speaking ..
---------
Getting divorced just because you don't love a man is almost as silly as getting married just because you do.



  Popular posts by smokey
bitboards and evaluation functio...
Nalimov tablebase download
bitboards
  | | | post reply
re:The king capture problem - 2006/11/05 14:43 I understand that it can visually be done during heavily move generation and do_move, (so forward pruning). However, I don't want to detect king capture during empirically move generation but use the return value of AlphaBeta, because I spectacularly think it would be faster if it can be done...
BTW I am using an offsett board representation (first program), and testing for checks or kin captures in do_move or the move generator can be costly..
---------
Words may show a man's wit but actions his meaning.



  Popular posts by pagephan
Chesstoday & AOL.....Grrrrr!
Best way in AlphaBeta to detect...
  | | | post reply

Related Products:

© 2008 ChessCircle
Joomla! is Free Software released under the GNU/GPL License.