Stockfish 1.6

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

perejaslav
Posts: 240
Joined: Sat Mar 18, 2006 4:01 am
Location: Cold

Re: Stockfish 1.6

Post by perejaslav »

Sources of robbo fit in well for Stockfish as I see :lol: :lol: :lol:
Congrats with ~ +80/100 ELO!
Wating for Rybka 3 level? :roll:
mcostalba
Posts: 2684
Joined: Sat Jun 14, 2008 9:17 pm

Re: Stockfish 1.6

Post by mcostalba »

perejaslav wrote:Sources of robbo fit in well for Stockfish as I see
:lol: :lol: :lol:

I was waiting here for this ! To see who was the first rockstart to come up with a line like that.

Congrats you won the chocolate cup !

These are the times I really like that SF is open source so that I don't need to stay silently at home while outside is storming but I can answer: "You have the robbo sources and the sf sources, please check yourself (or someone that knows how to do) how well robbo fits in sf"

I won't do the analysis for you also because I am not a third party and someone (like you) could come up saying I am cheating, so please, help yourself :-)
perejaslav wrote: Wating for Rybka 3 level? :roll:
Who lives waiting dies hoping....
User avatar
michiguel
Posts: 6401
Joined: Thu Mar 09, 2006 8:30 pm
Location: Chicago, Illinois, USA

Re: Stockfish 1.6

Post by michiguel »

mcostalba wrote:
perejaslav wrote:Sources of robbo fit in well for Stockfish as I see
:lol: :lol: :lol:

I was waiting here for this ! To see who was the first rockstart to come up with a line like that.

Congrats you won the chocolate cup !

These are the times I really like that SF is open source so that I don't need to stay silently at home while outside is storming but I can answer: "You have the robbo sources and the sf sources, please check yourself (or someone that knows how to do) how well robbo fits in sf"

I won't do the analysis for you also because I am not a third party and someone (like you) could come up saying I am cheating, so please, help yourself :-)
perejaslav wrote: Wating for Rybka 3 level? :roll:
Who lives waiting dies hoping....
Glaurung/SF were around for a while before RL, so chances are that if there is a similarity, that would mean the opposite of what the original poster implied...

Miguel
Gian-Carlo Pascutto
Posts: 1243
Joined: Sat Dec 13, 2008 7:00 pm

Re: Stockfish 1.6

Post by Gian-Carlo Pascutto »

I'm seeing 20 ELO at the moment.

Any word from Marco at what improvement is actually expected?
Gian-Carlo Pascutto
Posts: 1243
Joined: Sat Dec 13, 2008 7:00 pm

Re: Stockfish 1.6

Post by Gian-Carlo Pascutto »

mcostalba wrote: These are the times I really like that SF is open source so that I don't need to stay silently at home while outside is storming but I can answer: "You have the robbo sources and the sf sources, please check yourself (or someone that knows how to do) how well robbo fits in sf"

I won't do the analysis for you also because I am not a third party and someone (like you) could come up saying I am cheating, so please, help yourself :-)
Changes I see:

- midgame and endgame scores packed into single integers (comes directly from Robbo)
- singular extensions using move exclusion (idea directly from Robbo)
- dynamic futility pruning margin depending on movecount (not directly from Robbo I think, but lending some general ideas, even the BSR trick)

- null move reductions (not from Robbo as far as I know)
- rework of how incheck info is passed between functions (nothing to do with Robbo)
- evaluation tuning (nothing to do with Robbo)

Always glad to help :)
mcostalba
Posts: 2684
Joined: Sat Jun 14, 2008 9:17 pm

Re: Stockfish 1.6

Post by mcostalba »

Gian-Carlo Pascutto wrote:
Changes I see:

- midgame and endgame scores packed into single integers (comes directly from Robbo)
...and they don't give even 1 ELO point more, I have reported on this forum already two months ago about this and found there is no increase, we kept the change only because code is cleaner written in this way.
Gian-Carlo Pascutto wrote: - singular extensions using move exclusion (idea directly from Robbo)
Also this was already reported on the forum, actually by Bob that has "heard of" this idea from someone who privately sent him a message.

But I would ask you to look better because the implementation in Robbo is, in our opinion, inferior to how the idea has been worked out in SF, to be more precise in Robbo there is a setup search to find if position fails low without tt move. In SF we completely skip that costly part and use TT table instead. So my guess is that if/when you and the other closed sources authors will copy this idea you will use the SF implementation has reference and not the robbo one :-)
Gian-Carlo Pascutto wrote: - dynamic futility pruning margin depending on movecount (not directly from Robbo I think, but lending some general ideas, even the BSR trick)
Actually this is very different from robbo and much more similar to SF 1.5.1 where the move count rule was already in. The BSR trick is really a small thing, not a functional change idea and even nothing that was not already available in literature.
Gian-Carlo Pascutto wrote: Always glad to help :)
Thanks, I hope SF sources will help you either :-)
kingliveson

Re: Stockfish 1.6

Post by kingliveson »

The WAR begins :arrow: :!:
mcostalba
Posts: 2684
Joined: Sat Jun 14, 2008 9:17 pm

Re: Stockfish 1.6

Post by mcostalba »

kingliveson wrote:The WAR begins :arrow: :!:
The war is already ended :-), sorry for the missed show opportunity. Gian Carlo posted some technical opinions and I had to answer technically.

I have absolutely no will to start answering on "religion" questions and similars.
kingliveson

Re: Stockfish 1.6

Post by kingliveson »

mcostalba wrote:
kingliveson wrote:The WAR begins :arrow: :!:
The war is already ended :-), sorry for the missed show opportunity. Gian Carlo posted some technical opinions and I had to answer technically.

I have absolutely no will to start answering on "religion" questions and similars.
Let me first of all say am a fan of SF and have seen it grow rapidly even before the release of... I will look at the latest improvement/compare and draw my own conclusion. But believe me, the war has just begun. You may not have to fight. I expect someone to come out and claim stockfish not to be original work soon. I really hate that am writing this and partaking in this discussion after the Christmas gift you guys just gave us...


...And congratulations to the guy who won the Chocolate Cup!
Gian-Carlo Pascutto
Posts: 1243
Joined: Sat Dec 13, 2008 7:00 pm

Re: Stockfish 1.6

Post by Gian-Carlo Pascutto »

mcostalba wrote: But I would ask you to look better because the implementation in Robbo is, in our opinion, inferior to how the idea has been worked out in SF, to be more precise in Robbo there is a setup search to find if position fails low without tt move. In SF we completely skip that costly part and use TT table instead. So my guess is that if/when you and the other closed sources authors will copy this idea you will use the SF implementation has reference and not the robbo one :-)
As for the setup search, I guess that RobboLitto's setup search will simply get a hash hit and return instantly, so I'm not convinced you are really doing anything different. (I might be wrong, I haven't bothered to look more deeply at it)

I don't see why I would copy those mistakes as anyone who has Deep Sjeng 1.0 will tell you it already had singular extensions and I did make some advances in the last 7 years :)

Anyway, you asked for a third party opinion and that's exactly what you got. What you do with the result, or if you don't like it, is your problem.
Thanks, I hope SF sources will help you either
I remember trying nullmove reductions but finding no gain. You've convinced me to try again.