When will we see HOUDINI in official tournaments?

[Harvey Williamson]

I am not interested in mudslinging. Have a nice weekend.

So why did you start it? This was a civilised discussion about rules and their implementation until you came in!

Bye
Ingo

+1

[michiguel]

...
Maybe a hybrid system in which every engine is sim-tested and the ones who do not pass it get an automatic investigation would be better?

Miguel

That sounds like a reasonable compromise, just how to satisfactory, plausible test a remote engine - and assuming you can, there is no reliable data on clusters regarding the similarity ... Catch 22!

So far, only accused engines have been investigated, which always has stricken me as unfair.

If this is unfair the question is if it is better to test the ones who are on site but let some play without testing as you are not able to test them ...?

Bye
Ingo

Of course not, but that is my point, cluster and remotes could easily be tested. Certainly easier than the current system.

Miguel

[IWB]

...
Maybe a hybrid system in which every engine is sim-tested and the ones who do not pass it get an automatic investigation would be better?

Miguel

That sounds like a reasonable compromise, just how to satisfactory, plausible test a remote engine - and assuming you can, there is no reliable data on clusters regarding the similarity ... Catch 22!

So far, only accused engines have been investigated, which always has stricken me as unfair.

If this is unfair the question is if it is better to test the ones who are on site but let some play without testing as you are not able to test them ...?

Bye
Ingo

Of course not, but that is my point, cluster and remotes could easily be tested. Certainly easier than the current system.

Miguel

That is the point where we dissagree. If they could be tested is one thing but I doubt that they will!

Regards
Ingo

[hgm]

Ingo, Robolito is an Ippolit clone, thus not allowed. If an engine gives just one 60% hit with any other engine that engine is not allowed to play.

I see no reason for that. If an engine does have 90% similaity with an engine that does not participate, why shouldn't it play? I just wouldn't want both to play.

The test should be limited to participants only. Otherwise it would not be feasible. For one, every engine will have older versions of itself with high similarity score, so no engine would be allowed to participate anymore. In the second place, you would have to test participants against a potentially infinite supply of engines, some of them you might not even have access to.

So Robbolito is an Ippo clone, and the similarity score can show it. But if Ippolit isn't participating, I see no reason why that should prevent Robbolito's participation. It can represent the Ippo group, and it would be interesting to have a representant of that group in the tourney.

[S.Taylor]

By the way, i am very thankful to Vas for the interest and intrigue his great cluster would be to me, if there is top competition with it.
I am very much looking forward to seeing what a super super chess entity can do, if it is way above a pc Houdini 3, in its strength of play, even if it mostly due to the extremely powerful hardware.
So i feel this is actually a big service to chess.

[Cubeman]

I am looking forward to the Leiden too, to those competitors that don't compete because they think that since rules can't be enforced or don't trust your opponents to play fair,then I feel sad for you.

[geots]

............But it also makes sense that tournaments would want entrants that do not always pick the same moves.

-Sam

Indeed! Makes it much more interesting from a spectator's point of view.

My solution would be to let everybody play on a giant all play all basis and then the warring factions (evangelists, left anti-capitalist, devil worshippers, technocrats, the bored, Hyatt fans, Vas fans, Semites and anti-Semites) can battle it out as to which of the many possible results cross tables is "gods own cross table". Politically incorrect cross tables can be sent to a new secret forum UCT (unmentionable cross tables).

Yea, but then you are going to have at least one ordained priest and 2 licensed exorcists present. The exorcists are in case Hyatt and Levy show up.


gts

[Sedat Canbaz]

I am sure the CSVN can provide a good operator for you. The more relevant issue is if your current engine stays below the required 60% similarity, which is already pretty tolerant.

Hello dear Ed,

Just my 2 cents over this issue

First of all,i am happy that soon there will be a new ICT 2012 Leiden
Many thanks to all participants and to the tournament operators,managers too
And there is no doubt that ICT 2012 Leiden will a very exiting and interesting

About Rybka Cluster participation,
No patience to see what will be the performance of this monster
Honestly, i expect Rybka will be the winner in this great tourney

About the rules:60%
Good idea....but however nowadays its hard to find 100% original work
That's why Fire,Vitruvius,Strelka,Ivanhoe... should be allowed to participate too
Actually 100 % original program is that program which plays different strong modern openings
*Note:currently almost all chess engines play/repeat almost same openings
(I mean in case of,when the engines are playing without using an opening book)

About ICT 2012 Leiden hardwares,which will be used
I agree with the current hardware conditions and as far as i know all official tournaments are based on unlimited speed hardwares
In case of not allowing processor speed advantage,that means the opening books should be limited too
*Note:not only the processor speed,the opening book is another very important factor and plays a lot of role for high chess engine performance
But however,i don't see any reason of such hardware speed or opening book limitations
Because,nowadays we have a lot of sources,played on same hardware and using same opening book
Thats why...a such chess competitions (unlimited books/hardwares) will be always a great fun and interest

Btw,it's a sad news that we can't see Houdini,Komodo,Fruit and plus some previous world champions too (in the current tour):
Junior
Hiarcs
Shredder
Crafty
Chess Tiger
Rebel
Fritz
vs...


Best,
Sedat

[Mike S.]

This actually supports my doubts. The method spots eval similarities but does not spot, to a sufficient degree, search differences. 100ms vs. 400ms still give ~73%. That means, if someone takes Fruit 2.1's evaluation but makes the search code four times more effective (leading to your 400ms results in 100ms), this achievement would not be recognized sufficiently by the testing method.

This is a problem.

I insist that a similarity analysis must not ignore differences of the search code. It cannot be based on eval only, because reasonable eval terms are within a limited horizon. It is no surprise that they are similar among very good engines and only small details are different.

[Uri Blass]

This actually supports my doubts. The method spots eval similarities but does not spot, to a sufficient degree, search differences. 100ms vs. 400ms still give ~73%. That means, if someone takes Fruit 2.1's evaluation but makes the search code four times more effective (leading to your 400ms results in 100ms), this achievement would not be recognized sufficiently by the testing method.

This is a problem.

I insist that a similarity analysis must not ignore differences of the search code. It cannot be based on eval only, because reasonable eval terms are within a limited horizon. It is no surprise that they are similar among very good engines and only small details are different.

If this is the case then we can expect to see big similiarity between the best programs but it is not the case.

Note that if you make the program 4 times faster then not only the evaluation is constant but also the order of moves is constant(and it is important if you start from move A or start from move B in the search to decide if you pick move A or pick move B).

I think that if you modify another program having a rule that you change your mind in the first iterations if you get better or equal evaluation instead of changing your mind only in case of better evaluation is going to reduce similiarity between engines and I wonder how much.