The secret of Houdini

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

Moderators: Harvey Williamson, bob, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Guenther
Posts: 3436
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: The secret of Houdini

Post by Guenther » Wed May 16, 2012 8:04 pm

Jimmy Huggins wrote:Maybe Osipov Jury should get Dimitri Yasenko for copy right infringement on his user name --> Strelkaman :lol:

Makes as much sense has his argument.
Yeah its troll season again. They always start with an agony of posts...
which is a good indicator.

Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 10:58 am
Location: Antalya/Turkey
Contact:

Re: The secret of Houdini

Post by Sedat Canbaz » Wed May 16, 2012 9:22 pm

Zdrastvui Dimitri,
Sedat is right.
Thank you...i am happy that you noticed what i mean
The question that you have to do is why Sedat chose to change the priorities settings?I
Since more than two years i am running the engines with priority below normal
(I am using this useful priority recommendation by H.G.Muller)
Sedat make engine tests with Ponder OFF
Yes...its true that my latest rating list is based on with Ponder OFF

Please see here why i switched to running matches on same PCs (with Ponder OFF):
http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=24866

But note also that i am running Auto232 (Ponder ON) matches almost 10 years

Plus i have thousands of games (with Ponder ON) played by Houdini and in my every testing: Houdini is ranking 1st

Still i am waiting a recommendation about in which setting i need to test the Houdini's opponents
I mean,where Houdini will start to loose

In case of such priority recommendation:then i promise to run a Auto232 test against each other between two PCs

Then really i wonder a lot about what will be Opponent's Elo difference against Houdini ?!

Btw,the below priority data is captured during the engines thinking process:
Image
Image


Image
Image


Image
Image



Best,
Sedat

User avatar
Strelkaman
Posts: 83
Joined: Tue May 15, 2012 11:38 am

Re: The secret of Houdini

Post by Strelkaman » Wed May 16, 2012 11:02 pm

Sedat Canbaz wrote:Zdrastvui Dimitri,
Sedat is right.
Thank you...i am happy that you noticed what i mean
The question that you have to do is why Sedat chose to change the priorities settings?I
Since more than two years i am running the engines with priority below normal
(I am using this useful priority recommendation by H.G.Muller)
Sedat make engine tests with Ponder OFF
Yes...its true that my latest rating list is based on with Ponder OFF

Please see here why i switched to running matches on same PCs (with Ponder OFF):
http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=24866

But note also that i am running Auto232 (Ponder ON) matches almost 10 years

Plus i have thousands of games (with Ponder ON) played by Houdini and in my every testing: Houdini is ranking 1st

Still i am waiting a recommendation about in which setting i need to test the Houdini's opponents
I mean,where Houdini will start to loose

In case of such priority recommendation:then i promise to run a Auto232 test against each other between two PCs

Then really i wonder a lot about what will be Opponent's Elo difference against Houdini ?!

Btw,the below priority data is captured during the engines thinking process:
Image
Image


Image
Image


Image
Image



Best,
Sedat
It was a misunderstanding,Sedat.Iam very sorry.I thought that you wrote "above normal",but you wrote "below normal".Keep on your great work.
Strelka is the best engine in the world!!

User avatar
geots
Posts: 4790
Joined: Fri Mar 10, 2006 11:42 pm

Re: The secret of Polichinelle

Post by geots » Thu May 17, 2012 2:36 am

hgm wrote:
Tennison wrote:1. I don't say it's the case with Houdini but it's very easy to hide something to the task manager ... For example : the viruses ... They don't appear in task manager but are still running.
That is because virusses by nature hide in existing programs, and are not CPU intensive tasks. If they would use CPU time, you would see it in the task manager. If only because they eat the time away from the process you are monitoring, so that it shows less than 100%.

2. A know example of Process Thief is Ivanhoe 58 ...

Try it with ponder off and you'll see it freeze the other engine and it is always running !

Maybe Houdini can do like Ivanhoe 58 but more slightly (without completely freezing the other engine but just slow down a little) and maybe Houdini can also have an hidden process running ...
Sure. It is possible for an engine to disobey the ponder-off command, use more CPUs than you tell it to use, use more hash memory than you tell it to use... I know some Xiangqi engines that steal CPU. But you always find out pretty easily with the task manager that something fishy is going on.
You said 8ply for Houdini is not 8ply for Critter. Can you explain ?
Because for me it seems the same ...

If I test 7 ply / 8 ply (the 8 win nearly everything), and if I test 8 ply / 7 ply (the 8 win nearly everything) so the only comparable is with the same ply ...
Have you ever tested TSCP against Stockfish or Houdini at 8 ply? That would be a very educative experience! The difference between top engines, which all have very similar searches, is of course only modest, but when you try something really different...


The Russians wanted Fischer searched because they believed he had some kind of transmitter on him that was interfering with Spaasky's thinking. :roll:



gts

Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 10:58 am
Location: Antalya/Turkey
Contact:

Re: The secret of Houdini

Post by Sedat Canbaz » Thu May 17, 2012 6:24 am

Strelkaman wrote:
It was a misunderstanding,Sedat.Iam very sorry.I thought that you wrote "above normal",but you wrote "below normal".Keep on your great work.
Hello dear Dimitri,

No problem... its ok

Btw,do you have any info about Strelka Mp-when will be released ?

Greetings,
Sedat

User avatar
Strelkaman
Posts: 83
Joined: Tue May 15, 2012 11:38 am

Re: The secret of Houdini

Post by Strelkaman » Thu May 17, 2012 8:24 am

Sedat Canbaz wrote:
Strelkaman wrote:
It was a misunderstanding,Sedat.Iam very sorry.I thought that you wrote "above normal",but you wrote "below normal".Keep on your great work.
Hello dear Dimitri,

No problem... its ok

Btw,do you have any info about Strelka Mp-when will be released ?

Greetings,
Sedat
Soon.In less than a month.I guess:)
Strelka is the best engine in the world!!

User avatar
ATOMICC
Posts: 150
Joined: Sat Mar 10, 2012 10:50 pm
Location: USA
Contact:

Re: The secret of Houdini

Post by ATOMICC » Thu May 17, 2012 11:58 am

Entertaining thread.
Happy chessing!

Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 10:58 am
Location: Antalya/Turkey
Contact:

Re: The secret of Houdini

Post by Sedat Canbaz » Thu May 17, 2012 8:21 pm

Strelkaman wrote:
Sedat Canbaz wrote:
Strelkaman wrote:
It was a misunderstanding,Sedat.Iam very sorry.I thought that you wrote "above normal",but you wrote "below normal".Keep on your great work.
Hello dear Dimitri,

No problem... its ok

Btw,do you have any info about Strelka Mp-when will be released ?

Greetings,
Sedat
Soon.In less than a month.I guess:)
Good news...

Probably i will wait (for SCCT) the release of Strelka MP version


Best,
Sedat

User avatar
geots
Posts: 4790
Joined: Fri Mar 10, 2006 11:42 pm

Re: The secret of Houdini

Post by geots » Fri May 18, 2012 2:55 am

Strelkaman wrote:
Tennison wrote:To avoid this problem when testing, I think it's simple : just test with same depth and not with a time control.

I just try a fast test on 100 games at 8 Ply depth for each engine ...

Better is to test with at least depth 20 but too much time necessary for this (I can only do these long test during the night).

Code: Select all

Gui = Fritz 11
Book = HS 7 moves + (ctg)
Hash = 64mb
                        
1   Critter 1.4 64-bit    +40/-18/=42 61.00%   61.0/100  (+78 elo)
2   Houdini 2.0b Pro x64  +18/-40/=42 39.00%   39.0/100  (-78 elo)
And yes it seems very strange ... but need more games and more depth to conclude anything ... If someone can test it ...

INTERESTING approach by Tennison.Let me think it for a while.What Tennison did is that he locked the plies for both engines.Houdini can't maximize his abilities because he can't use the CPU memory through the advance of plies.My theory is that Houdini absorbs CPU processor memory abilities faster than any other engine.Thats why the processor is overheating faster when Houdini is running.Well Done Tennison.Iam looking forward to see other tests from you


Nice to meet you George Speight.



And nice to meet you as well, Dimitri. Your immense knowledge is well known in certain circles. Glad to see you here and hope you stay.


All the best,

george

Post Reply