Page 2 of 2

Re: Plausible move generator

Posted: Sat Nov 10, 2012 7:48 pm
by syzygy
diep wrote:That's software progress. Yet i do not believe the search is the biggest reason for that.
Yes, evaluation is certainly one of the reasons Stockfish at relatively low nps still does so well. But also at low nps selectivity in the search clearly helps a lot, when done well.
Please note that the manufacturers aren't serving us well lately. The ivy bridge nonsense you can buy in the shops now has just 4 cores. Where is the hardware progress?
There are now reports that IB-E processors coming out next year will only have 6 cores like SB-E. This is disappointing. At 22nm it should be possible to place up to 16 IB-cores on a die. Haswell will probably not improve on this either.

Re: Plausible move generator

Posted: Sat Nov 10, 2012 9:17 pm
by Jorge Garcia
Yes, Vincent, it implies not to search all the moves, as you said it's a Shannon A approach, I did not know about Paradise so I will take a look
at the code.
Thanks for the link.
Best Regards

Re: Plausible move generator

Posted: Mon Nov 12, 2012 6:18 pm
by diep
Jorge Garcia wrote:Yes, Vincent, it implies not to search all the moves, as you said it's a Shannon A approach, I did not know about Paradise so I will take a look
at the code.
Thanks for the link.
Best Regards
If you're gonna write chess patterns and rules, why not directly improve the evaluation function from Toga and win elo instead of lose 2000 elo?

Re: Plausible move generator

Posted: Wed Nov 14, 2012 1:45 pm
by Jorge Garcia
Hi Vincent,
I am trying to write a move generator that discards bad moves at the same
level, this is in what I am involved now. I know that this concept is strange and dangerous, but I like to experiment new things...

Re: Plausible move generator

Posted: Wed Nov 14, 2012 4:04 pm
by diep
Jorge Garcia wrote:Hi Vincent,
I am trying to write a move generator that discards bad moves at the same
level, this is in what I am involved now. I know that this concept is strange and dangerous, but I like to experiment new things...
If someone pays i wouldn't complain if i were you.

Re: Plausible move generator

Posted: Wed Nov 14, 2012 4:12 pm
by diep
Jorge Garcia wrote:Yes, Vincent, it implies not to search all the moves, as you said it's a Shannon A approach, I did not know about Paradise so I will take a look
at the code.
Thanks for the link.
Best Regards
Note that Shannon A searches all moves. Shannon B just a few.

Re: Plausible move generator

Posted: Fri Nov 16, 2012 3:27 pm
by Jorge Garcia
Thanks Vincent,
I am digging into this idea for some months ago and until now I only find
blunders, and nothing at all. I know that it's like loosing your spare time...
Best regards :P :P