Stockfish 1.8 JA available

[alpha123]

1.8 is probably only about +20 ELO, but nevertheless it was a good idea to publish a new version. Bugs were fixed, redundant code was removed (for example queen contact mate detection in eval is gone!), many things were reorganized, the executables are noticable faster now. From what I've seen so far, I'm very happy with the 1.8 release. Good work! Germany vs. Argentina was ok, but most of the time I prefered to look at the 1.8 sources.

A 20 Elo improvement at the top is quite substantial, you guys have some unrealistic expectations. Congrats to the Stockfish team for continuing to improve their great engine!
And Germany was awesome this afternoon, made Argentina look like a very poor team (like England) .

Robert

20 elo in 3-ish (I think; when was 1.7 released?) months is very good. Plus cleaned up codebase, fixed bugs, etc. Amazing effort, Stockfish team.

Yeah, I think someone's been poisoning the South American water. First Brazil losing to the Netherlands (of all places), now Argentina losing 4-zilch to Germany.

Peter

[gerold]

Logo by David Dahlem
Stockfish 1.8 JA (UCI)
by Tord Romstad, Marco Costalba & Joona Kiiski.
http://www.stockfishchess.com/

Windows/Linux 64/32 Intel p.g.o compiles

Download:
http://www.mediafire.com/?dnyvmf4vtaw
Mirror:
http://good.net/_uTVWSc3qY

Jim.

Thanks Jim and Stockfish team.
Appears to be just a llittle better than Rybka 3.
Maybe a 20+ elo gain.

Best,
Gerold.

[Eelco de Groot]

^^ 1.8 is probably only about +20 ELO, but nevertheless it was a good idea to publish a new version. Bugs were fixed, redundant code was removed (for example queen contact mate detection in eval is gone!), many things were reorganized, the executables are noticable faster now. From what I've seen so far, I'm very happy with the 1.8 release. Good work! Germany vs. Argentina was ok, but most of the time I prefered to look at the 1.8 sources.

Hello Ralph,

I haven't gotten as far as you in studying all the changes in the code but nevertheless already diligently working on Rainbow Serpent 1.8s (dc) Just for fun, not a real effort to make a fork, not this time.

If 1.8 would be plus 20 elo, it will pass Rybka 3 on Ingo's list, that would be quite something already. But Frank Quisinsky thinks it may actually be less in his testing so we are not sure yet. I have not had a chance to look at search and eval yet but I am systematically changing every mention of "Stockfish" into "Rainbow Serpent 1.8s(dc)" so watch out for a new smooth scaling version of Stockfish one of these days It will probably be much weaker than Stockfish 1.8.

Maybe too much has changed now to easily import some of the "smooth scaling ideas" into the code. And some things are going to be superfluous, e.g. for "blocked pawns" there is now "opposed pawns", in Joona's version I think, and the pawn storm code is simplified, I have not yet decided if I will leave that intact or change it into something else again. Search will undoubtedly be the most unrecognizable bit although I really have not looked at that yet, I take your word for that and Marco said there is no separate PV search anymore, at least that was his idea, a while ago. I think I will have to do that differently though, if that really is the case.

Regards, Eelco

[Nimzovik]

Frankly.... I am just thrilled to live in this 'Golden Age of Chess.' The new engines that appear so rapidly- a far cry from Rex chess and Computer Chess Reports. Stockfish....Houdini...... Rybka 4 ..... Firebird..... Hiarcs..............Shredder......... Fritz.......Chessmaster.........!! Not to mention the Great Guis- Arena , Aquarium Chessbase. Free database programs like Scid and chess base light. Many internet chess Gui's to choose from for FICS. Free internet chess servers. Internet chess on a grand scale. I remember when ICC had 800 players at it's peak!! Great forum. Downloadable opening books every where!! Chess blogs! You tube chess videos!! And of course the debate about Pablo anti - chess!!! Who - just who can really righteously complain and be taken seriously Zippity do dah, zippity ay...My o my what a wonderful day...... YEEEEEEEEEEEEEEEEEEHHHHHHHAAAAAAWWWWWWWWWWW

[gerold]

[mcostalba]

I am systematically changing every mention of "Stockfish" into "Rainbow Serpent 1.8s(dc)" so watch out for a new smooth scaling version of Stockfish one of these days It will probably be much weaker than Stockfish 1.8.

Let people to play with updated and very changed sources (although functionality is more or less the same of 1.7.1) is the reason why we pushed this version, also if we knew the difference with previous one to be small in terms of ELO.

Realistically I expect SF 1.8 to be on par with R3, i.e. within elo error bar for the various lists. This would be already an important result for us considering that R3 was the strongest in the world (apart fom Ippo derivates) until about one month ago. This would mean for us to end a long catch-up run started 2 years ago, when SF was light years from R3 and this result of today (that people even complain as too small ) was deemed a "mission impossible".

Probably on Frank's list SF 1.8 will end up below 1.7.1, but there I would guess IMHO that was 1.7.1 to be a bit overrated given that was well above R3. Now SF 1.8 will end up on R3 levels where it has to be.

[Frank Quisinsky]

Hi Marco,

SF 1.7.1 played 1.200 games (overrated not possible). For a stable rating 650 games are enough +-10. Only in 1/52 cases a rating will have more as +-10 if 650 games played ... if enough opponents in a list.

With more time Rybka 3 lost more ELO compare to the others. SWCR is over factor 2.0 more time as IPON.

Interesting too that SF 1.8.0 lost a lot of more points vs. weaker engines ... compare to version 1.7.1.

Furthermore, Rybka 3 x64 is around 50 ELO stronger as Rybka 3 w32. Stockfish around 30 ELO (w32 to x64). At the moment I test the 32-Bit version.

Now around 450 games played. Same results for SF 1.8.0 after 100, 150, 200, 250 ... up to 450 games.

Very interesting at all for me too ... because I am not at home and looking on the actual LIVE tournament table

3 games won SF 1.8.0 up to mate under 55 moves after 450 games.
More as 35 games won SF 1.7.1 after 800 games up to mate under 55 moves. That's what I mean ... where is the middlegame power from version 1.7.1?

Best
Frank

[mcostalba]

Hi Marco,

SF 1.7.1 played 1.200 games (overrated not possible). For a stable rating 650 games are enough +-10. Only in 1/52 cases a rating will have more as +-10 if 650 games played.

With more time Rybka 3 lost more ELO compare to the others. SWCR is over factor 2.0 more time as IPON. I think its possible that SF 1.8.0 lost more points as v. 1.7.1 to the others with more time.

Interesting too that SF 1.8.0 lost a lot of more points vs. weaker engines ... compare to version 1.7.1.

Furthermore Rybka 3 x64 is around 50 ELO stronger as Rybka 3 w32. Stockfish around 30 ELO. At the moment I test the 32-Bit version.

Now around 450 games played. Same results for SF 1.8.0 after 100, 150, 200, 250 ... up to 450 games. Very interesting at all for me too ... because I am not at home and looking on the actual LIVE tournament table.

3 games won SF 1.8.0 up to mate under 55 moves after 450 games.
More as 35 games won SF 1.7.1 after 800 games up to mate under 55 moves. That's what I mean ... where is the power from version 1.7.1?

Best
Frank

Hi Frank,

sorry, I didn't mean to say something wrong. For sure your list is correct. Regarding your question "where is the power from version 1.7.1" I don't know the answer

[Frank Quisinsky]

Hi Marco,

it can be a random but if you look in the table ...

SF 1.8.0 made 2.800 - 2.850 ELO only vs. engines which played very aggressive middlegames.

(Hiarcs, Junior, Spark, Thinker, Twisted).

SF 1.8.0 made more points vs. engines which played a strong endgame, late middlegame (Rybka, Critter, Komodo for examples).

Complete other build have SF 1.7.1.

Vs. Hiarcs, Junior, Spark, Thinker, Twisted ... SF 1.7.1 made clearly more points. After around 450 games I don't think it is random but let us wait. Monday in the evening I have the complete results.

Best
Frank

I sayed it yesterday in the other thread. After all I saw the endgame is again better but in the middlegame SF 1.8.0 lost agressiveness. With more and more games this problem seems to be more clear.

Or I see a ghost or have a statistic problems.
My brain have a problem, to warm ... other things
All is possible

[mcostalba]

Hi Marco,

it can be a random but if you look in the table ...

SF 1.8.0 made 2.800 - 2.850 ELO only vs. engines which played very aggressive middlegames.

(Hiarcs, Junior, Spark, Thinker, Twisted).

SF 1.8.0 made more points vs. engines which played a strong endgame, late middlegame (Rybka, Critter, Komodo for examples).

Complete other build have SF 1.7.1.

Vs. Hiarcs, Junior, Spark, Thinker, Twisted ... SF 1.7.1 made clearly more points. After around 450 games I don't think it is random but let us wait. Monday in the evening I have the complete results.

Best
Frank

I sayed it yesterday in the other thread. After all I saw the endgame is again better but in the middlegame SF 1.8.0 lost agressiveness. With more and more games this problem seems to be more clear.

Or I see a ghost or have a statistic problems.
My brain have a problem, to warm ... other things
All is possible

Hi Frank, I have sent to you an e-mail.