Search found 469 matches

by rvida
Sat Jan 12, 2013 11:36 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: All and Cut nodes
Replies: 69
Views: 14277

Re: All and Cut nodes

lucasart wrote:For example, does it work to have a different min depth and reduction for IID between 'All' and 'Cut' nodes
In 'All' nodes you can omit IID entirely since all moves are expected to fail low.
by rvida
Sat Jan 05, 2013 2:31 pm
Forum: Computer Chess Club: General Topics
Topic: Bug in Critter-1.6a?
Replies: 4
Views: 1071

Re: Bug in Critter-1.6a?

This happens from time to time. With SMP search some hash entries get corrupted due to concurrent writes.
by rvida
Sat Dec 29, 2012 2:55 am
Forum: Computer Chess Club: General Topics
Topic: some discussion of GUI's
Replies: 55
Views: 4495

Re: some discussion of GUI's

Matthias Gemuh wrote:
Don wrote:...
May I ask which programming language / environment you are using for this?
ancient Borland C++ Builder 5.
ouch!
by rvida
Mon Dec 24, 2012 9:12 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Adjustable search pruning depending on time control
Replies: 34
Views: 5458

Re: Adjustable search pruning depending on time control

Yes, I quite forgot about that little detail. We couldn't find any benefit from this idea of varying reduction with depth when in check. Even reducing in check at all is rather a close call. I'm pretty sure that if the above code has any benefit, it would be worth a fraction of an elo point (compar...
by rvida
Mon Dec 24, 2012 1:44 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Adjustable search pruning depending on time control
Replies: 34
Views: 5458

Re: Adjustable search pruning depending on time control

As far as I know, Rybka, Ippo/Ivanhoe, Critter, Houdini (at least 1.5, don't know about 3.0) and all other engines of which I have knowledge agree with Komodo on this point. Not entirely true. They do some depth based reductions - in cut nodes when evading check. The following code snippet is from ...
by rvida
Fri Dec 14, 2012 5:33 pm
Forum: Computer Chess Club: General Topics
Topic: Elo points gain from doubling time
Replies: 78
Views: 22908

Re: Elo points gain from doubling time

I did a similar partial study for Critter 1.2 and 1.6, to see if the pattern is much different. As far as I know, Critter 1.2 does not have any search feature other than Singular Extension that kicks in at 9 ply, so this would appear to fully account for the observed jump in Critter Easy to verify....
by rvida
Fri Dec 14, 2012 2:18 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: program style, risk aversion
Replies: 83
Views: 8393

Re: program style, risk aversion

Don wrote: I doubt that 1 point has any impact
Contempt=1 does not mean draw score of 1 point. It is actually 6 to 20 points, depending on game phase (measured at root).

Setting contempt=2 does same as contempt=1 plus adds some king safety asymmetry.
by rvida
Fri Dec 14, 2012 2:11 pm
Forum: Computer Chess Club: General Topics
Topic: Elo points gain from doubling time
Replies: 78
Views: 22908

Re: Elo points gain from doubling time

I did a similar partial study for Critter 1.2 and 1.6, to see if the pattern is much different. As far as I know, Critter 1.2 does not have any search feature other than Singular Extension that kicks in at 9 ply, so this would appear to fully account for the observed jump in Critter Easy to verify....
by rvida
Fri Dec 07, 2012 3:20 am
Forum: Computer Chess Club: General Topics
Topic: New distributed project:Release date for well known engines
Replies: 19
Views: 2426

Re: New distributed project:Release date for well known engi

Critter 0.34 2009-04 Critter 0.38 2009-05 Critter 0.39a 2009-06 Critter 0.42 2009-10 Critter 0.52 2010-01 Critter 0.52b 2010-02 Critter 0.60 2010-04 Critter 0.70 2010-05 Critter 0.80 2010-07 Critter 0.90 2010-11 Critter 1.0 2011-03 Critter 1.01 2011-03 Critter 1.2 2011-06 Critter 1.4 2011-12 Critte...
by rvida
Tue Nov 27, 2012 7:52 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: SMP and questions
Replies: 68
Views: 11047

Re: SMP and questions

You might be able to reduce the search overhead a bit by updating alpha to a value stored in a location shared by all threads searching the particular PV node, each time before you start a search Makes sense. I thought everyone is doing that. If not, they should. or re-search of a move in a PV node...