Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Houidini 3 at slower time controls
1 2 3 4 5 Previous Next  
Parent - By Morpheus (****) Date 2012-09-25 13:10
NO,  + BENKO ;-)
Ok Good Luck , thank You for info.
Parent - - By elemeNt Date 2012-09-26 16:39
Houdini largely based on stolen code.
Parent - By Christian Packi (****) Date 2012-09-26 17:20
Nice first post in 4 years. :yell:
Parent - By gotogo (**) Date 2012-09-26 17:39
I guess that can be said for Rybka stolen code from fruit... let it rest clown.
Parent - - By Nelson Hernandez (Gold) Date 2012-09-26 18:00
I took my granddaughter to a children's zoo last weekend.  There was a goat pen there.  Your post reminds me of how it sounded.  Behhhhh!  Behhhh!  Made my granddaughter start crying.
Parent - - By Barnard (Bronze) Date 2012-09-26 18:03
you are grand-father?you are relatively young Nelson...if i didnt miss something
Parent - - By Nelson Hernandez (Gold) Date 2012-09-26 19:31
56 next month.  Plenty old to be a grandfather!
Parent - - By Barnard (Bronze) Date 2012-09-27 04:23
only 56?you are a young man...and that is good,you will have too much time to enjouy of your grand-son/s :smile:
Parent - - By Nelson Hernandez (Gold) Date 2012-09-27 04:45
Hah!  I only have one grandchild, but I just found out a few hours ago a second one is on the way!  Pop the champagne bottles, maybe this one will be a grandson!
Parent - - By Barnard (Bronze) Date 2012-09-27 04:47

> I just found out a few hours ago a second one is on the way!


congratulations Nelson!

you want to have a grandson to be named like you? :wink:
Parent - - By Nelson Hernandez (Gold) Date 2012-09-27 12:11
Nelson III would be most welcome!
Parent - By Barnard (Bronze) Date 2012-09-27 12:13
that sounds good,making live the good names trough the generations :smile:
Parent - By Moz (****) Date 2012-09-29 02:55
Congratulations!  I have a son attending UC Santa Barbara and I half expect to receive the very same news every time he calls!
Parent - - By Geomusic (*****) Date 2012-09-27 04:31
You are going to be 56? I thought you were like 49 or 50...
Parent - - By Nelson Hernandez (Gold) Date 2012-09-27 04:46
The main thing I have learned: you will never be younger than you are right now.
Parent - By Geomusic (*****) Date 2012-09-27 04:54
I am learning this slowly at 38...
Parent - By Highendman (****) Date 2012-09-26 22:11
:)
Parent - - By Pia (****) Date 2012-10-02 08:02
You call people "goats" and still not even an apology?!
While too lazy to look at, I'll do it for you:

Left: houdini 1.0, Right: "Robart" 0085g3


Text "Pawn at rank one or eight" just replaced by "Pawn at rank 1 or 8" and so alike.

Also, with all this,

(From start position, 128 MB hash, Fritz 11 GUI)

Analysis by Houdini 1.01 w32 1_CPU: 2010

1.Ng1-f3 Nb8-c6 2.Nb1-c3 Ng8-f6 3.d2-d4 d7-d6 4.h2-h3 Bc8-e6 5.e2-e4
  =  (0.03)   Depth: 6/16   00:00:00  1kN
1.e2-e4 e7-e5 2.Ng1-f3 Ng8-f6 3.Nb1-c3 Nb8-c6 4.Bf1-c4 Bf8-c5 5.0-0 0-0 6.d2-d3 d7-d6 7.Nc3-a4 Bc8-e6 8.Na4xc5 d6xc5 9.Bc4xe6 f7xe6 10.Bc1-e3 Qd8-d6 11.c2-c3 Ra8-d8
  =  (0.06)   Depth: 19/40   00:00:18  17838kN
1.Nb1-c3 Ng8-f6 2.d2-d4 d7-d5 3.Bc1-f4 c7-c5 4.e2-e3 c5xd4 5.e3xd4 a7-a6 6.Ng1-f3 Nb8-c6 7.Bf1-e2 e7-e6 8.0-0 Bf8-d6 9.Qd1-d2 0-0 10.Nf3-e5 Qd8-b6 11.Ne5xc6 Qb6xc6 12.Bf4-e5 Bd6xe5 13.d4xe5
  =  (0.06)   Depth: 23/52   00:05:23  313mN

Analysis by RobboLito 0.085g3 w32: 2009

1.Ng1-f3 Nb8-c6 2.Nb1-c3 Ng8-f6 3.d2-d4 d7-d6 4.h2-h3 Bc8-e6 5.e2-e4
  =  (0.06)   Depth: 6/21   00:00:00  2kN
1.e2-e4 e7-e5 2.Ng1-f3 Nb8-c6 3.Bf1-c4 Ng8-f6 4.Nb1-c3 Bf8-c5 5.d2-d3 0-0 6.0-0 d7-d6 7.Nc3-a4 Bc8-e6 8.Na4xc5 d6xc5 9.Bc4xe6 f7xe6 10.Bc1-e3 Qd8-d6 11.c2-c3 Ra8-d8
  =  (0.10)   Depth: 18/42   00:00:14  11737kN
1.d2-d4 Ng8-f6 2.Nb1-c3 d7-d5 3.Bc1-f4 c7-c5 4.e2-e3 c5xd4 5.e3xd4 a7-a6 6.Ng1-f3 Nb8-c6 7.Bf1-e2 e7-e6 8.0-0 Bf8-d6 9.Qd1-d2 0-0 10.Nf3-e5 Qd8-c7 11.Qd2-e3 Bc8-d7 12.Ne5xd7 Bd6xf4 13.Nd7xf6+ g7xf6
  =  (0.13)   Depth: 22/52   00:06:37  318mN
Parent - - By Homayoun_Sohrabi_M.D. (***) Date 2012-10-02 09:44
Please inform the Robbo author about your findings and let him take appropriate action against Mr Houdart.   Do not buy the product if you think it's stolen and I think that's probably where your concern should end.
Parent - - By Pia (****) Date 2012-10-02 10:31
Not sure how. "Yakov Petrovich Golyadkin" is a fictional name from Dostoevsky's novel. Also Igor Igorovich Igoronov, Roberto Pescatore or Jon Doe...
Parent - - By Homayoun_Sohrabi_M.D. (***) Date 2012-10-02 11:49
Exactly!   You claim it's "stolen" code.   How can you steal code that doesn't even have an owner?   So you are saying that Houdart took some code that has no owner, that is freely available to all on the internet and he improved on it ....and you call that stealing? 
If you want to criticize somebody, you can criticize the Robbo guys, but I don't believe Houdart has done anything wrong.    If it makes you feel better, eventually his code will be RE'ed too and somebody will make a better engine based on the Houdini framework.
Parent - By Pia (****) Date 2012-10-02 12:31
So, one anonymous e-mail from that guy under fictional name and Houdini
would be shot down? Somehow Houdart knows it would never happen...
Parent - - By Nelson Hernandez (Gold) Date 2012-10-02 13:21
I didn't call anyone a goat.  Please read more carefully.  I only made a connection between what he said and the sour bleating of the goats I had witnessed and had upset my granddaughter.  I stand by that.  No apologies.
Parent - By Pia (****) Date 2012-10-02 16:05
Okay, I'm sure it was an accident, something beyond control.
Parent - By Pia (****) Date 2012-10-01 12:10

>Houdini largely based on stolen code.


No, it just slightly modified Robbolito. While they have very similar lines in analysis,
file sizes also much, suggesting that they both may have same source files.
Parent - - By RubiksChess (**) Date 2012-10-01 12:23
Stolen code from an engine that is largely based on stolen code.

Turns out code is stolen a lot, by your definition. Turns out that as people figure out better algorithms, other people tend to use them to improve their work as well.

Turns out you may not have a point at all.
Parent - - By Pia (****) Date 2012-10-02 09:55 Edited 2012-10-02 10:04
Isn't kranium and sentinel banned with their programs as cloners? Why houdini is different?

Steal once and you're thief. Steal twice and you're respectful owner.
Parent - By RubiksChess (**) Date 2012-10-02 11:02
Quite simply because the world of computer chess has become politically divided and corrupt.

Now you can get by on "street cred" and can get drowned in BS and biases just because you're in the opposing camp.

A lot of people measure the health and success of the computer chess world by the ELO gaps of the new contenders. I measure it by how friendly it is to competition.

By that standard, the current state of the computer chess world is dismal at best.
Parent - - By Stonehenge (***) Date 2012-09-26 22:53
The second match just finished.

Final result of the Houdini 3 - Stockfish 2.3.1 match: +44 -16 =60
74-46 (+82 Elo ± 42 Elo).
Download Games

Final result of the Houdini 3 - Houdini 2.0c match: +48 -15 =57
76.5-43.5 (+94 Elo ± 42 Elo).
Download Games

The results of both matches have very large confidence intervals, please consider these when discussing results.
My impression is that Houdini 3 had a very favorable run against Houdini 2.0c and a slightly unfavorable run against Stockfish 2.3.1.
Neither engine scored 40% against Houdini 3, let's see whether Komodo 5 fares any better :).

Houdini 3 - Komodo 5 match has now started - no games finished yet.
Parent - - By Christian Packi (****) Date 2012-09-26 22:58
What's the performance if you combine both matches?
Parent - - By Stonehenge (***) Date 2012-09-26 23:10

> What's the performance if you combine both matches?


So far Houdini 3 has performed at (+88 Elo ± 30 Elo) compared to the average of Stockfish 2.3.1 and Houdini 2.0c.
Difficult to say anything more accurate, the rating difference between Stockfish 2.3.1 and Houdini 2.0c at the match conditions is unknown.
Parent - By Christian Packi (****) Date 2012-09-26 23:13
That’s fine with me.

Btw would you make a Mac OS version of Houdini if I say Please? :smile:
Parent - - By Werner (***) Date 2012-09-27 07:27 Edited 2012-09-27 08:42
My results on CEGT conditons for 40/20 list are:
                         
1   Houdini 2.0c x64 1CPU +56     +17/-9/=24 58.00%   29.0/50
2   Stockfish 2.3.1 x64 1CPU-56   +9/-17/=24 42.00%   21.0/50


I have got a similar result for Houdini 1.5a

Here are Ingos results:
                       
1   Houdini 2.0 STD +51     +54/-32/=64 57.33%   86.0/150
2   Stockfish 2.3.1 JA +51  +32/-54/=64 42.67%   64.0/150


Werner
Parent - - By Stonehenge (***) Date 2012-09-27 11:12
Werner, thank you.
If I'm not mistaken this match is at about 3 times longer effective time control (taking into account the hardware) than CEGT 40/20.
So it's closer to CEGT 40/120 than to CEGT 40/20.
Parent - By Werner (***) Date 2012-09-27 11:38
Yes, of course:
Your match is much longer - and close to CEGT 40/120!

Werner
Parent - - By FWCC (***) Date 2012-09-27 13:36
What do you attribute this great jump in strength to?Is it added chess knowledge,faster scalability,tricks with the weight settings or what?At this pace Houdini will/should remain number 1 for a long time or until another breakout engine or if an existing one improves by leaps and bounds(Rybka?)
Parent - - By Stonehenge (***) Date 2012-09-27 14:37

> What do you attribute this great jump in strength to?Is it added chess knowledge,faster scalability,tricks with the weight settings or what?At this pace Houdini will/should remain number 1 for a long time or until another breakout engine or if an existing one improves by leaps and bounds(Rybka?)


There are many improvements and new ideas in all areas of the engine, the approx. 50 Elo gain is the accumulation of a lot of small steps.
Contrary to the popular belief on chess forums there's not a lot to be gained by fiddling with weight settings, the only way forward is by adding knowledge and new techniques to the engine.
Parent - - By Carl Bicknell (*****) Date 2012-09-27 15:19

> Contrary to the popular belief on chess forums there's not a lot to be gained by fiddling with weight settings, the only way forward is by adding knowledge and new techniques to the engine.


Is Houdini 3's superior positional evaluation (c/p Houdini 2) due to search refinements allowing deeper search in quiet positions, or to adding 'chess knowledge'? (it seems to be the trend NOT to do that these days & to take out chunks of code)
Parent - By Stonehenge (***) Date 2012-09-27 22:18

> Is Houdini 3's superior positional evaluation (c/p Houdini 2) due to search refinements allowing deeper search in quiet positions, or to adding 'chess knowledge'? (it seems to be the trend NOT to do that these days & to take out chunks of code)


Both. Houdini's progress in every version has always been a combination of improved evaluation (more and better knowledge) and better search (smarter move picking, better selectivity).
The Houdini 3 evaluation is in fact a lot more expensive than in Houdini 1.03. The fact that the node speed has never visibly declined between versions is simply because of always better, hand-optimized code.
Parent - - By Razor (****) Date 2012-09-27 18:56
Sorry to jump in mid-way through a conversation with others, I just wanted to put a question to you.

Now that the Houdini development is delivering products that your customers can buy, do you intend to take feedback from the same as suggestions on where the development focus should be going.  If the answer to this is yes then will you open up an appropriate section on your website to allow/encourage feedback?
Parent - By Stonehenge (***) Date 2012-09-27 22:21

> Now that the Houdini development is delivering products that your customers can buy, do you intend to take feedback from the same as suggestions on where the development focus should be going. If the answer to this is yes then will you open up an appropriate section on your website to allow/encourage feedback?


We accept feedback via e-mail, our address is shown on the web page.
As in all software projects, only a small number of suggestions will actually make it into the product, but any feedback is appreciated.
Parent - - By Nelson Hernandez (Gold) Date 2012-09-26 23:41
Not a single instance of Stockfish dominating both sides of the board from these test positions.
Parent - - By Gaмßito (****) Date 2012-09-27 02:11

>Not a single instance of Stockfish dominating both sides of the board from these test positions.


No. It did in the Marshall Attack, in game 11 and 12. Anyway the final result is quite good for Houdini.

Regards,
Gaмßito.
Parent - By Nelson Hernandez (Gold) Date 2012-09-27 04:44
I stand corrected.  It's hard to tell whether the final result is good for Houdini as we don't have a solid sense of Stockfish 2.3.1 yet.  I may be a bit behind the curve on that, though.

The Marshall game where Stockfish is white is an endgame loss for Houdini, a universal weakness in chess engines.  I could be wrong but it seems black can still manage a draw after black's 56th move:

8/1k6/7p/8/PK2r3/8/1R5P/8 w - - 0 57


Someone ought to put this one to Victor.
Parent - By Gaмßito (****) Date 2012-09-27 02:30

> My impression is that Houdini 3 had a very favorable run against Houdini 2.0c and a slightly unfavorable run against Stockfish 2.3.1.


Yes. And I have seen that in the match against Stockfish 2.3.1, Stockfish has always used much more time [at least until move 40] than Houdini.
Perhaps still the time management with increment can be slightly improved in H3.

Regards,
Gaмßito.
Parent - By Geomusic (*****) Date 2012-09-27 04:32
yay pgns of the strongest chess games played on earth...well at blitz lvl anyway :)

not including the cluster of course :) It plays standard level games at 1 move per sec LOL
Parent - - By Stonehenge (***) Date 2012-09-27 09:31

> Houdini 3 - Komodo 5 match has now started - no games finished yet.


Current standing after 72 games: +26 -12 =34

Parent - - By Highendman (****) Date 2012-09-27 15:04
This is so awesome. I haven't been this excited by computer chess for two years! Thanks, Robert!
Parent - - By Nelson Hernandez (Gold) Date 2012-09-27 15:37
It's ironic, Robert is using Vas's own website to promote his product, and Vas is all right with that, apparently!  As it should be!

This is the best possible revenge for all the travails that have taken place over the past two years: superior products in the pipeline!  Here's hoping Rybka 5 is comparable or better, Komodo MP is a sensation, Stockfish and Critter continue to climb, and all this activity triggers a new golden age of computer chess in 2013...and NONE of them participate in ICGA tournaments!  How satisfying that would be.
Parent - By Highendman (****) Date 2012-09-27 15:45
I agree on the ICGA part 100%.

However on the timeframe for other developers catching up with Houdini (even after the reverse engineering they will all do once its out) - I don't see it happening in the next 18 months.

Every not so often a breakthrough hard working engineer / genius comes around... And it takes the rest of the pack a vey long time to catch up.
Up Topic The Rybka Lounge / Computer Chess / Houidini 3 at slower time controls
1 2 3 4 5 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill