Stockfish 1.7

[mcostalba]

Could you (and also all the other people that experience problems with CPU detection) please post type of CPU you have ? Thanks.

Hi Marco,

same problem here on my Quad Q6600, Win7 64bit.
Stockfish 1.7 only shows Threads=2 by default.

Joerg

Thanks to you and to all other people reporting this. I think it is a problem with i7 CPU, so we very probably will revert the new HT detection code and turn back to 1.6.3 behaviour.

In the meantime you can manually set "threads" UCI parameter to 4 from your preferred GUI.

[James Constance]

You are correct to say I am using an Atom processor. Hope it can be fixed!!

Is it possible for you to compile yourself ?

To verify if the problem is in the binary targeted to x86 CPU

Hi Marco - I'd be happy to have a go if someone would explain what to do! But it's not something I've done before outside of "Basic4ppc".

James

If you have some experience with compilers you can download MSVC express (it is free):

http://www.microsoft.com/express/Downlo ... Visual-CPP

then from File->new->project from exsisting code

You could try to point to the SF sorces directory and let MSVC read the files, then compile (there is the corresponding menu).

I have to warn you that you _could_ waste some time in the effort, theoretically, once you have download and installed MSVC it is more or less a 2 minutes effort...but only if you have already done this before.

Thanks for the pointer, Marco. I'll let you know if I make any progress.

[IWB]

Hi Marco

... we don't expect a big ELO increase from 1.6.X series (if any),..

That is the best understatement I have read for a long time!

Big Thx 4 the engine!

Bye
Ingo

[mcostalba]

Hi Marco

... we don't expect a big ELO increase from 1.6.X series (if any),..

That is the best understatement I have read for a long time!

Big Thx 4 the engine!

Bye
Ingo

Honestly we don't have a clue how was this possible

The possibility that cross my mind is that the JA builds gave an unexpected push to ELO. Indeed we do all our test with self compiled versions and only the final one is sent to Jim to prepare the release.

Could be that the JA compile this time is really super fast, much faster then ours.

[Dr.Wael Deeb]

Hi Marco

... we don't expect a big ELO increase from 1.6.X series (if any),..

That is the best understatement I have read for a long time!

Big Thx 4 the engine!

Bye
Ingo

Honestly we don't have a clue how was this possible

The possibility that cross my mind is that the JA builds gave an unexpected push to ELO. Indeed we do all our test with self compiled versions and only the final one is sent to Jim to prepare the release.

Could be that the JA compile this time is really super fast, much faster then ours.

Ya ya ya ya,it's the bloody compile that added the leap in the Elo
Come on Marco,don't be that humble
Cheers,
Dr.D

P.S.No offence is intended toward master Jim of course

[IWB]

Hi

Honestly we don't have a clue how was this possible

It is your task to find out how, I am just happy 'that'!

You know what the biggest accomplishment is? Yesterday, 8th of April at 19.50 Server time, for the first time in 5 years Rybka was not the most loaded engine on Playchess! Congratulations!

If my current test continues like this it might be that Stockfish becomes the new No.1!

Bye
Ingo

[Tord Romstad]

Hi Marco

... we don't expect a big ELO increase from 1.6.X series (if any),..

That is the best understatement I have read for a long time!

It was no understatement -- we did not expect an improvement anywhere close to this. I suppose we are better programmers than testers. Still, I think Stockfish will drop a bit on your list when more games are played, and also that it won't score quite as well on the various other rating lists. It seems that your playing conditions are somehow (unintentionally, of course) very well optimized for Stockfish.

By the way, thank you very much for runng these tests! It's been very exciting to follow your results; I think I must pressed the "refresh" button on my web browser on average about once every ten minutes today. I'm glad you don't offer live transfer of the actual games while they are being played. If you did, I wouldn't have been able to do any work at all today.

[Uri Blass]

Hi Marco

... we don't expect a big ELO increase from 1.6.X series (if any),..

That is the best understatement I have read for a long time!

It was no understatement -- we did not expect an improvement anywhere close to this. I suppose we are better programmers than testers. Still, I think Stockfish will drop a bit on your list when more games are played, and also that it won't score quite as well on the various other rating lists. It seems that your playing conditions are somehow (unintentionally, of course) very well optimized for Stockfish.

By the way, thank you very much for runng these tests! It's been very exciting to follow your results; I think I must pressed the "refresh" button on my web browser on average about once every ten minutes today. I'm glad you don't offer live transfer of the actual games while they are being played. If you did, I wouldn't have been able to do any work at all today.

It seems the opposite

stockfish performs very well against the weak engines so far and if it is going to continue with the same performance it is going to be the new number 1.

Note that poor spike is losing 20-0 against it and this result is not included in the performance calculations.

Stockfish 1.7 JA - Hiarcs 12.1 MP 32b (2657) 15.5 - 5.5 73.81% Perf=2836
Stockfish 1.7 JA - Critter 0.52b (2645) 17.5 - 2.5 87.50% Perf=2983
Stockfish 1.7 JA - Deep Junior 11.1a (2634) 18.0 - 3.0 85.71% Perf=2945
Stockfish 1.7 JA - Loop 13.5/6 (2606) 19.5 - 1.5 92.86% Perf=3051
Stockfish 1.7 JA - Twisted Logic 20100131x (2593) 19.0 - 2.0 90.48% Perf=2984
Stockfish 1.7 JA - Spike 1.2 Turin 32b (2574) 20.0 - 0.0 100.00%
Stockfish 1.7 JA - Crafty 23.1 JA (2540) 19.0 - 1.0 95.00% Perf=3051
128.5 - 15.5 89.24% Perf=2950




144 out of 700 games played

[IWB]

I suppose we are better programmers than testers.

Just assuming this is true, you have to worry about it as there are possibilities that you might have thrown away some really good improvements!!!

It seems that your playing conditions are somehow (unintentionally, of course) very well optimized for Stockfish.

That might be a possibility, but how likely is that? All engines fit more or less into the other lists and now, just for Stockfish 1.7, my conditions fit very very well? If so, it is interesting in itself!

By the way, thank you very much for runng these tests!

Thx for the thx, it is nice to see that someone is interested!

It's been very exciting to follow your results; I think I must pressed the "refresh" button on my web browser on average about once every ten minutes today.

Me too I find this always very exciting and becasue of that I am doing this for years, I just thought it might be interesting for others as well - and this seems to be right! Unfortunately I am a minimalistic HTML programmer (euphemistic!) and my site by far is not sophisticated enough (but at least no bells and whistles!).

I'm glad you don't offer live transfer of the actual games while they are being played. If you did, I wouldn't have been able to do any work at all today.

I could, but then you would have to follow 8 games at once ...
(and the worst thing is: I can see the games ... )

Bye
Ingo

[Zagalo]

...

Honestly we don't have a clue how was this possible

The possibility that cross my mind is that the JA builds gave an unexpected push to ELO. Indeed we do all our test with self compiled versions and only the final one is sent to Jim to prepare the release.

Could be that the JA compile this time is really super fast, much faster then ours.

Make a test between your compile and JA's and give us the difference

If JA's compile is 15+ELO better then Jim Ablett should charge for this! and we might just see Rybka-JA soon