Houdini 1.5 is released

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

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
UncombedCoconut
Posts: 319
Joined: Fri Dec 18, 2009 10:40 am
Location: Naperville, IL

Re: Houdini 1.5 is released

Post by UncombedCoconut » Sat Dec 18, 2010 9:00 am

Graham Banks wrote:
Peter Skinner wrote:On another forum (The Ippolit forums), even they see it as a clone.
Plus there's this post there:

the size of houdini compiles points towards the mini nalimove end game bases being included ,they are extremly small but highly effective , one of the winiboard GUIs has them built into it, and would account for its end game knowledge

it may account for Robert Houdunits reluctance to honour his word to show his source code ....:)

any doubters such bases or GUI exists may like to see

Portable setup of Winboard 4.5b(master build 07.08.2010.)

interstingly ...guess whoose engine is also included in package ?
I think that's unlikely on technical grounds:
  • Testers like Ingo allow EGTBs in other programs, but still see a strength increment.
  • The engine already supports GTBs, which are at least as easy to embed.
  • The executable size is consistent with Robert's asserted (and probably insane ;)) method of supporting multiple CPUs, which is by duplicating search code. So is his choice to release separate executables depending on the number of CPUs in prior versions.
Furthermore, since the Houdini executables are uncompressed, it's irresponsible to make such an allegation without proof. I'm not merely rehashing the "don't assert your conscience without proofs!" argument -- it's dead easy to confirm or falsify such a claim using a common substring search. So why not do it? (Of course, one could embed obfuscated Nalimov bases in the program -- but that's more trouble than using a legal method!)

User avatar
Houdini
Posts: 1471
Joined: Mon Mar 15, 2010 11:00 pm
Contact:

Re: Houdini 1.5 is released

Post by Houdini » Sat Dec 18, 2010 9:42 am

lucasart wrote: Thanks. This is great news! But is there any chance for a Linux or Mac OS version ?

If you have a look at the source code of StockFish, you'll see it's a matter of a few #ifdef's for thread functions, either using the Windows or POSIX ones, and adding a gcc line in the makefile. Or you can contact Jim Ablett for help with compiling on Linux and also Mac OS. I think he uses the Intel compiler, which is somewhat better than gcc.
Lucas, there will be no Linux or Mac OS release in the near future, my limited time resources simply don't allow this.
I am keeping an open mind on this matter, my position can change anytime in the future.

User avatar
Houdini
Posts: 1471
Joined: Mon Mar 15, 2010 11:00 pm
Contact:

Re: Houdini 1.5 is released

Post by Houdini » Sat Dec 18, 2010 9:49 am

Peter Skinner wrote:Where is the source?

If you took ideas (and likely code) from Crafty, Stockfish, and the Ippolit family of engines, you have to publish the code.

Crafty's license states that any modification of the code _must_ be released and the copyright notice be left intact.

I would like to see just how much of Crafty is in the program.

Peter
Houdini contains some ideas but zero code from Crafty.

Here's one example.
Like Crafty, Houdini will reinitialize the TT entries when getting close to the 50-move rule limit in order to have the correct evaluations taking into account the 50-move draws.

Cheers,
Robert

User avatar
Houdini
Posts: 1471
Joined: Mon Mar 15, 2010 11:00 pm
Contact:

Re: Houdini 1.5 is released

Post by Houdini » Sat Dec 18, 2010 10:05 am

Graham Banks wrote:Plus there's this post there:

the size of houdini compiles points towards the mini nalimove end game bases being included ,they are extremly small but highly effective , one of the winiboard GUIs has them built into it, and would account for its end game knowledge
LOL. That is such idle speculation.
Houdini doesn't contain any built-in endgame table other than the KPK ending.
Any dissambler will easily verify that Houdini contains specific knowledge for a number of endings. For example, Houdini is probably the only engine that implements the Averbakh rules in KBP v KB endings.

Robert

User avatar
Houdini
Posts: 1471
Joined: Mon Mar 15, 2010 11:00 pm
Contact:

Re: Houdini 1.5 is released

Post by Houdini » Sat Dec 18, 2010 10:20 am

tmokonen wrote:I tried it out on my old PIII Windows 2000 Server SP4 box, and I got the same error message. Are you using VC++ 2010? I've read on Microsoft's support site that W2K is not a supported platform for executables made with that compiler.
Thanks for the info!
I am using VS 2010, and indeed it seems that the executable images are not compatible with Win 2000. From the MS web site:
Unfortunately, we deprecated support for building images that run against Windows 200 with the VS2010 version of the C\C++ libraries and compiler.
Conclusion is that Houdini will not run on Win2000.

User avatar
Mike S.
Posts: 1480
Joined: Thu Mar 09, 2006 4:33 am

Re: Houdini 1.5 is released

Post by Mike S. » Sat Dec 18, 2010 3:58 pm

ernest wrote: But why do you keep crippling yourself with things of the past (like W2K)?
This is due to difficult circumstances. :roll: Normally I'd of course like to have an i7 system and 64 bit Windows 7. I noticed that the 64 bit Windows seems to be standard now, on preconfigured comps which are being sold. - Nevertheless I cannot complain currently in terms of internet access and the common good freeware I used to have (e.g. Chess Hero), basically. I never did top engine competition nor corr.chess anyway.
By the way, have you been thinking about an update of your PB5moves.ctg (which I am still using in most of my engine matches)?
No... I have lost my large games databases. But I think the PB5moves.ctg is still entirely ok for the intended purpose.

For anyone else who'd like to have it: I don't find the files anymore, but you can achieve the same if you use the Balanced-12.ctg or the Balanced-16.ctg and limit the book depth in the book settings in the Fritz interface. These two books can be downloaded from:

http://members.aon.at/computerschach/li ... #downloads

(The Balanced-16 is just a deeper version from the same database as the Balanced-12. See readme.txt for details.)
Regards, Mike

Robert Flesher
Posts: 1246
Joined: Tue Aug 18, 2009 1:06 am

Re: Houdini 1.5 is released

Post by Robert Flesher » Sat Dec 18, 2010 4:33 pm

Peter Skinner wrote:
Roger Brown wrote: Hello Peter Skinner,

I am afraid that you might need to add me to that list of fools.

Ideas do not equal code unless I have missed something.

Demanding that code be released seems to presume that you or any member of the Crafty team has a right to make such demands. That right would have to come from some basis apart from wanting to satisfy your desire to know, correct?

As for the TD position, I agree with you here. Any TD may enforce their rules, including and excluding any engine they desire. Nothing wrong with that. However you are sadly mistaken if you think that the buzz around this engine has one iota to do with whether it will be allowed to compete at some event or not.

All of the buzz is speaking to the strength of the thing, that it is abusing the top of the commercial and free engine food chain.

I am neither an apologist or a mouthpiece but it irks considerably that demands are made in this case with nothing more than your desire to satisfy your need to know how much - which means you currently have no idea - Crafty code is in the engine whereas in other cases (with at least some [inconclusive?] evidence) these demands for code were at best muted.

Sigh.

Calling it a clone when the [invalid] argument of the defenders of that other engine was "this is the strongest there is so it benefited little from engine X which was there before and which it now outdistances" seems odd to say the least.

I do not accept the sheer strength argument BUT if it was advanced and accepted as a decisive reason why engine X could not have come from engine Y, then why are you denying that something original may have been created here as well?

I would like to know the truth (whatever that means) too but your demands seem unreasonable at the present moment given what we do not know.

And no, I have never written a chess engine....or even compiled one.

Later.
It is a clone. Do you honestly believe he took "ideas" from Ippolit, Crafty, Stockfish and wrote a 3200 engine from scratch in a few months?

Please tell me you aren't that blind to the fact of what Houdini is? If so, I have a hell of a deal on a bridge with a prime location in San Fransisco. Some would call it a "Golden" opportunity. Interested?

Peter

If it is a clone, and to this we have seen NO evidence but only conjecture. The idea of Houdini < Rybka is clearly "Argumentum ad Novitatem" It is likely that he did exactly what VAS did. Looked at code (Vas ,used Fruit, Robert / Ippo*) saw the ideas, and made them better. If it's ok for Vas, it should be fine for Robert. If it is only a clone why is it so much stronger ?

From a customer point of view, ( I paided good money for Rybka 2, 3, 4), Robert provides a stonger, better analysis tool, far better endgame player, and regular updates. All for free! If tommorrow he was to sell the next version, I'd buy it.

Lastly, Houdini appears to be at very least 50 ELO stronger than Rybka 4.0. Give credit where credit is due! As envy is the pursuit of fools, and you do not strike my as such.

Terry McCracken
Posts: 16464
Joined: Wed Aug 01, 2007 2:16 am
Location: Canada

Re: Houdini 1.5 is released

Post by Terry McCracken » Sat Dec 18, 2010 4:58 pm

Albert Silver wrote:
Nimzovik wrote:Thanks Mr. Houdart. Great engine.

Yikes. Mcaffee toolbar gives last posted site for download as questionable. I will wait till better site is provided.

Major :shock: Paranoid :shock: regards - :?
Just so you know, McAfee (together with Norton) is one of the record holders in false alerts.
I haven't any false alerts with Norton 2011.
Terry McCracken

User avatar
Peter Skinner
Posts: 1760
Joined: Sun Feb 26, 2006 12:49 pm
Location: Edmonton, Alberta, Canada
Full name: Peter Skinner
Contact:

Re: Houdini 1.5 is released

Post by Peter Skinner » Sat Dec 18, 2010 5:18 pm

Robert Flesher wrote: If it is a clone, and to this we have seen NO evidence but only conjecture. The idea of Houdini < Rybka is clearly "Argumentum ad Novitatem" It is likely that he did exactly what VAS did. Looked at code (Vas ,used Fruit, Robert / Ippo*) saw the ideas, and made them better. If it's ok for Vas, it should be fine for Robert. If it is only a clone why is it so much stronger ?

From a customer point of view, ( I paided good money for Rybka 2, 3, 4), Robert provides a stonger, better analysis tool, far better endgame player, and regular updates. All for free! If tommorrow he was to sell the next version, I'd buy it.

Lastly, Houdini appears to be at very least 50 ELO stronger than Rybka 4.0. Give credit where credit is due! As envy is the pursuit of fools, and you do not strike my as such.
What I am saying it is a clone of Ippolit, nothing more, nothing less. Has it improved on the original source. Sure. Am I saying people can't use it? Obviously no, as that is your choice, just as it is mine to not use it or any of the other derivatives in that family of engines.

Nor am I envious of what he has accomplished. Being on the Crafty team has taught me that even 1-5 elo gains are impressive, because the hard work involved and the insane amount of testing that goes into each of those small incremental increases feels awesome when they are verified improvements.

Taking someone's code, improving on it slightly when you already have one of the strongest source projects to start as a base doesn't impress me. Write something from scratch that beats the #1 program by 50 elo and I will be impressed.

My initial question was to see the source based on Robert Houdart's claim that he used "ideas" from Crafty. He states there is no actual code, and I have nothing to refute that with as there is no source to look at. So I have to take his word on that until one day it might be released.

In my opinion it is a clone, and I'm entitled to that opinion just like you are yours.

Peter
I got kicked out of Chapters because I moved all the Bible's to the fiction section.

Albert Silver
Posts: 2969
Joined: Wed Mar 08, 2006 8:57 pm
Location: Rio de Janeiro, Brazil

Re: Houdini 1.5 is released

Post by Albert Silver » Sat Dec 18, 2010 5:39 pm

Terry McCracken wrote:
Albert Silver wrote:
Nimzovik wrote:Thanks Mr. Houdart. Great engine.

Yikes. Mcaffee toolbar gives last posted site for download as questionable. I will wait till better site is provided.

Major :shock: Paranoid :shock: regards - :?
Just so you know, McAfee (together with Norton) is one of the record holders in false alerts.
I haven't any false alerts with Norton 2011.
I misquoted. Norton was not one of the top false alerts. McAfee is in the middle with a high number (three times more than Norton for example), but the kings of false alerts among well-known AVs are Panda and Kaspersky.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."

Post Reply