Search found 1154 matches

by Gian-Carlo Pascutto
Thu Mar 12, 2009 3:07 pm
Forum: Computer Chess Club: General Topics
Topic: Clustering etc. thread
Replies: 159
Views: 28314

Re: An idea for a new WCCC format - what do you think?

No it isn't. There is a big difference between searching deeply to find the best move, and doing a shallow search terminated by a bunch of very fast games to try to predict the quality of that branch. This doesn't probe anywhere near as deeply as the original search the original search. You're effe...
by Gian-Carlo Pascutto
Thu Mar 12, 2009 2:26 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: LMR
Replies: 60
Views: 12169

Re: LMR

I've noticed from others reporting here that R3 is worse on tactical positions than previous versions, suggesting that extensions were reduced, not increased. Ah, I've seen the exact opposite claim. Must've been very scientific tests whose reports we've both been reading. That's all well and good. ...
by Gian-Carlo Pascutto
Thu Mar 12, 2009 1:51 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: PVS
Replies: 44
Views: 14176

Re: PVS

Fixed: for ( all moves) { make if ( bSearchPv ) { score = -pvSearch(-beta, -alpha, depth - 1); } else { score = -pvSearch(-alpha-1, -alpha, depth - 1); if ( score > alpha && score < beta ) score = -pvSearch(-beta, -alpha, depth - 1); // re-search }...
by Gian-Carlo Pascutto
Thu Mar 12, 2009 1:30 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: PVS
Replies: 44
Views: 14176

Re: PVS

So this in fact searches full-window (in pv nodes) as long as alpha is not raised. The code you posted is simply wrong, for the reason you stated. // in fail-soft ... && score < beta ) is common This also looks wrong to me. Why would you search if score >= beta? Just cut, doesn't matter if we're fa...
by Gian-Carlo Pascutto
Thu Mar 12, 2009 1:14 pm
Forum: Computer Chess Club: General Topics
Topic: Fruit vs. Toga poll
Replies: 89
Views: 12563

Re: Fruit vs. Toga poll

When you manage to take the strongest open-source engine and make it 100 points stronger, this would be a big achievement. IMO the less lines of code you change the smarter you actually are. 100 lines of code -> 100 elo improvement means what-1 elo for 1 line? Not bad eh? Why shouldn't be that enco...
by Gian-Carlo Pascutto
Thu Mar 12, 2009 1:09 pm
Forum: Computer Chess Club: General Topics
Topic: Fruit vs. Toga poll
Replies: 89
Views: 12563

Re: Fruit vs. Toga poll

If the Toga developers haven't contributed a single idea, then why is the latest Toga almost 100 elo stronger than Fruit 2.2.1? That seems to paint Fabien as inept (which he is far from being) because some amateurs could rewrite a few lines of code and get 100 elo improvement. The reasoning here is...
by Gian-Carlo Pascutto
Thu Mar 12, 2009 12:47 pm
Forum: Computer Chess Club: General Topics
Topic: Fruit vs. Toga poll
Replies: 89
Views: 12563

Re: Fruit vs. Toga poll

When my program ships with a license that says that you are free to do whatever you like with it (apart from making it proprietary), what right do I have to tell you later that you are not allowed to use it in some tournament? The license governs copyright and distribution. A third party (tournamen...
by Gian-Carlo Pascutto
Thu Mar 12, 2009 12:42 pm
Forum: Computer Chess Club: General Topics
Topic: Fruit vs. Toga poll
Replies: 89
Views: 12563

Re: Fruit vs. Toga poll

As far as tournaments go, all of the rules prohibited clones prior to the release of Fruit much less Toga. The only reason I could see the ICGA allowing cluster Toga in was to allow a scientific experiment on the clustering of a chess program. This is completely false! Clones have always been allow...
by Gian-Carlo Pascutto
Thu Mar 12, 2009 10:07 am
Forum: Computer Chess Club: General Topics
Topic: Clustering etc. thread
Replies: 159
Views: 28314

Re: An idea for a new WCCC format - what do you think?

In practice, sharing high-depth entries doesn't really change the situation that much. You want the full search state for whatever part of the tree you are searching or your performance will plummet. Try disabling the lowest-depth 98% of your hash writes in Sjeng and see what happens. Vas That's no...
by Gian-Carlo Pascutto
Thu Mar 12, 2009 9:59 am
Forum: Computer Chess Club: General Topics
Topic: Clustering etc. thread
Replies: 159
Views: 28314

Re: Results from UCT parallelization

What were the UCT speedups for tree splitting and multiple runs? It is quite possible that our serial searches are somehow too "structured" and not soft/stochastic enough. It is a little insane to build alpha-beta trees around 30-move main lines. Vas http://www.cs.unimaas.nl/g.chaslot/papers/parall...