Not logged inRybka Chess Community Forum
Forum RybkaChess Help Search Register Login
Up Topic The Rybka Lounge / Computer Chess / Houdini 3 info
- - By TheHug (Bronze) [us] Date 2012-03-01 22:28
25 elo+ over Houdini 2
45 elo+ over Houdini 1.5
Parent - - By NATIONAL12 (Gold) [gb] Date 2012-03-01 22:34
well as i think 1.5 and 2 are equal,the statement appears ridiculous.
Parent - By TheHug (Bronze) [us] Date 2012-03-01 22:34
Just posted what was written.
Parent - - By TheHug (Bronze) [us] Date 2012-03-01 22:37
He did say one thing on the 2nd page that caught my interest.

Robert-"As for Houdini 3, it will do quite well at long TC - some search algorithm improvements will have an increasing effect with higher search depths."
Parent - - By NATIONAL12 (Gold) [gb] Date 2012-03-01 22:42
it will do quite well at long TC

what on earth does that mean.he has only tested at blitz and most likely 60 second games.
Parent - - By Uly (Gold) [mx] Date 2012-03-01 22:54
I've always thought that engines waste a lot of time doing something wrong when set to analyze a position for several hours to high depth, the branching factor kicks in and the new 4 hours will be used to change just a single meaningless PV move deep on the line, most likely, and a change of root move can be forced faster if you exclude this main move from analysis.

If Houdart makes behavioral improvements for when the engine is left to analyze at high depth, it would give better analysis similar to if the user was interacting with it (not saying Houdart is going to do that, just saying it's possible.)
Parent - By TheHug (Bronze) [us] Date 2012-03-01 23:11
I guess he will have to expand a pond that statement to understand what he did with it.
Parent - - By NATIONAL12 (Gold) [gb] Date 2012-03-01 23:32
in late middlegame i have found changes to root move as late as D42(16 pieces still on the board) with Rybka,i know this depth sounds incredible and is a move you may ignore with interactive use or Jump AN as i used to call it.:smile:
Parent - - By TheHug (Bronze) [us] Date 2012-03-01 23:34

> Jump AN as i used to call it.


Yeah I bet you have a few ideas us 4 cores or less folks can't do. :smile:
Parent - By NATIONAL12 (Gold) [gb] Date 2012-03-01 23:44
4 core can do the same as me it just takes longer,in my forum games i could just as easily played with 4 core computer i had plenty of time and never got near the wire.Everyone wants to play so fast these days i guess thats a reflection on life.:smile:
Parent - By Werewolf (*****) [gb] Date 2012-03-02 00:09

> Robert-"As for Houdini 3, it will do quite well at long TC - some search algorithm improvements will have an increasing effect with higher search depths."


I wonder which 2 engine developers that comment is aimed at....
Parent - By Dr.Wael Deeb (***) [jo] Date 2012-03-02 08:44
Agreed,he's massively speculating :cool:
Dr.D
Parent - - By Arrière Pensée (Gold) Date 2012-03-02 01:23
But! Who are the "we" in  "We've started working on Houdini 3..."
Parent - - By Razor (****) [gb] Date 2012-03-02 07:52
Perhaps George Hardeen?
Parent - By Arrière Pensée (Gold) Date 2012-03-02 08:48
Houdini is an escape artist- Robert Houdart took Ip's out of the cellar and brought it into the mainstream by give it wheels to move an engine to run with.  Ip's was in a straightjacket! Does Robert Houdart contain multitudes?!:wink:

Parent - - By Barnard (Bronze) Date 2012-03-30 01:40
an use of the 'royal we'?
Parent - - By Arrière Pensée (Gold) Date 2012-03-30 02:38

> an use of the 'royal we'?


:wink:
Parent - By Barnard (Bronze) Date 2012-03-30 02:57
:yell:
- - By rocket (***) [se] Date 2012-03-02 08:51
Houdini 2.0c 64-bit 4CPU  3367 

   Houdini 2.0c 64-bit  3342

How can there be such little difference on 4 cores to one! does it scale horribly or what?
Parent - - By Dr.Wael Deeb (***) [jo] Date 2012-03-02 09:01
I think it's a sign of good scaling but I could be wrong :cool:
Dr.D
Parent - By Barnard (Bronze) Date 2012-03-30 01:45
a sign of good scaling is add around 40 elo points each time you double the cores(and if im not wrong,around 60 elo points each time you double the speed of the processors),and if the example that he mentioned is right,only added 25 elo points,when it must add around 80 elo points higher,so if those numbers are right,it scale really poor
Parent - By Barnard (Bronze) Date 2012-03-30 01:46
where you obtained that?in what list?and in a long time control?
Parent - - By kgburcham (***) [us] Date 2012-03-30 02:19
Houdini 2.0c 64-bit 4CPU  3367 
Houdini 2.0c 64-bit  3342
How can there be such little difference on 4 cores to one! does it scale horribly or what?

I have different results than you rocket.
Your single thread is probably correct.
This pc is 4x2.6gig, old 920cpu.

1 thread used
Houdini 2.0c Pro x64 (1024 MB)
24/53   0:59   +0.06 (149.770.782) 2521

4 processor(s) found
Houdini 2.0c Pro x64 (2048 MB)
26/62   0:55   +0.03 (482.923.387) 8631
Parent - - By Barnard (Bronze) Date 2012-03-30 02:30
Hi kgburcham

the '3367' and '3342' are elo points,but i doubt that your '2521' and '8631' are really elo points;maybe Houdini can play weak to 2521 elo points,but im really sure cant play at the level of 8631 elo points,so i think you confussed those numbers

regards

Salvador
Parent - - By kgburcham (***) [us] Date 2012-03-30 02:53
I didnt read above his post to see that it was elo.
The word scale made me think he was talking about kns.
Thanks Salvador
Parent - By Barnard (Bronze) Date 2012-03-30 02:55
not,with scale he means scaling doubling the cores,meaning the increment of the elo

you are wellcome :smile:

whats your name?so i will know how to call you...of you preffer that i call you kgburcham?
- By rocket (***) [se] Date 2012-03-30 13:08
those figures were from CCrl by Houdini
Up Topic The Rybka Lounge / Computer Chess / Houdini 3 info

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill