WCCC 2009

[bob]

We will just have to disagree. The 200+ pound heavyweights are going to wear out smaller boxers, whether they have more skill or not is irrelevant for the most part.

You got me sir, but who is the best when different classes exists? No boxing analyst will put right now a heavyweight in the mythical list of best pound for pound.

Relate to earlier ultimate fighting where grable expert Royce Gracie will embarass much heavier and stronger fighters, if the heavyweight could not land a punch....

Why do you suppose the "heavyweights" don't have any weight limit, while the lower classes have tight bounds? Is it because a 200 pound fighter hits much harder, and can absorb more punches from a lightweight, due purely to physics? The heavyweight (sometimes called 'open') class is at the top of the heap. The WCCC should represent that group of players, not the ones farther down the "power" totem-pole.

[diep]

In most sports there is THE WORLD CHAMPIONSHIP, and then a bunch of things with qualifiers (under 18 world championship, special olympics gold medal, whatever). We have 1 qualifier, "Computer", implying the computer must make all decisions unaided. As many people here say, it is unusual, misleading, and just plain wrong to call something a computer world championship with no other qualifiers, and then put lots of qualifiers. If they want hardware restrictions so that the only computers that can be used are home computers (such as 8 core restriction) they can call it the micro-computer championships.

It is a sad, sad thing for computer chess in my opinion if there is no true world computer championship though (where there are no restrictions other than all decisions must be made by the computer program). I also consider it somewhat ironic that the rational is to make it accessible to the masses, when the only x-world champion chess computer the masses know is "Deep Blue", which would be banned from this event, and probably would never have been developed if such restrictions had always been in place. Who knows what wonderful advances in computer chess will be missed by having no completive outlet spurring on advances in highly parallel chess computation.

-Sam

Where do you think computer chess would be today without chess 4.x, belle, hitech, deep thought/deep blue, etc? Those programs, plus others like later Slate versions, Cray Blitz, LaChex, Cube, Chaos, Awit, - the list goes on and on, created the public interest that kept interest alive. Now we have some that want to limit the hardware in some form or another so that everyone can have such a machine. Makes no sense.

There is a number of arguments in favour of unlimited hardware for the world championship. To drop just a few:

a) do you want software to get the maximum out of old hardware so say 3300 elo out of a 4 core machine and still is hardly above 3300 elo of course at 16 cores, or do you want to give us the chance to show up with software that scales well in elo at faster hardware (at which we can or cannot test?), say 4000 elo at 16 cores? (this is also the argumentation why testing at blitz doesn't work well)

b) do you want to see the miracles of mankind like 99% of the population, the BEST mankind can achieve?

I do.

The ones that do not belong on planet Ego.

c) AMD and intel have different number of cores. Reality is that non-sold intel cpu's of 2000 euro each, and a total system price months later of 10k euro got used in world championships 2009.

Yet intels nehalem 3.2Ghz and 3.4Ghz cpu's that got used (overclocks to 3.6-3.8Ghz @ 8 cores max using turboboost), is only obtainable by a few.

AMD's istanbul 2.4Ghz * 6 cores is not allowed in an incarnation of 12 cores. Though 12 cores * 2.4Ghz = 28.8Ghz

Versus intels nehalem 3.6Ghz * 8 = 28.8Ghz

So the ICGA limited the world champs to intel.
Later on they also mentionned hyperthreading would get allowed.

So 16 logical cores were allowed.

A cheap 16 core AMD 4 socket * quadcore box of $8000 is not allowed however.

At ebay you can get the 2.2Ghz quadcore cpu that works for 4 sockets for 900 dollar, that is for 4 cpu's. So that's 16 * 2.2 = 35.2Ghz

So that's a machine of $2500 in total, even if you have to buy all other components new.

By limiting the hardware to X cores you AVOID people from joining at cheap assembled hardware from ebay. That can be a 4 socket machine with a bit older cpu's, or a 8 socket machine with a bit older cpu's, or a cluster self assembled.

d) you piss off a lot of people by suddenly limiting the hardware whereas the world title as decided in 2001 would be open to any hardware. So the ICGA itself has changed in a meeting with everyone the championship to open hardware, now a single pencil brush modifies it back.

there weren't too many participants already. If you can't beat 16 core opponents you sure can't beat them at 8 cores either IN A LEGAL NON CHEATING MANNER.

e) it is a fairy tale that people are interested in limited stuff. They know all about limits already as they aren't Dagobert Duck themselves. They don't watch NASA when NASA toys with astronauts in kindergarten, they like to see the most unbelievable hardware the ultimum mankind can reach.

f) in past there was a zillion different world titles which all got sold to different manufacturers. $5000 for this title, $20k for that title (as joining with 4 machines) and so on. Big money for ICGA officials. The first incarnation of David Levy (the one that could play chess very well) was in those events also present i understood. Not sure about Jaap. Note i heard Jaap doesn't get paid a penny from ICGA money AFAIK and he uses up holiday time of the few weeks he has as a professor.

However nowadays the machines are way too strong. 3300 or 3400 elo, do YOU know the difference? I do not, but well i'm just a deteriorating FM of around 2300 FIDE.

There is that little participants that all these different world titles make no more sense. Sure i would have won the 'amateur title' by now several times if it would exist. So there is your proof that i'm not selfish.

How about the few in favour of the 8 core limit. It's just writing the big I in worlds largest font. Or in latin: EGO.

g) Counting cores is difficult. It is worlds most expensive cpu's that are high clocked. Getting 1000 low clocked slow cores by definition is a lot cheaper than getting 4 high clocked ones. ICGA never had a limit on Ghz limit.

If you really want to limit the hardware, limit the number of Ghz,
by taking the square root out of the number of cores times the MAXIMUM Ghz frequency of them (where they can clock themselves automatically at during search - note this is again confusing with intel).

The most expensive cpu always is a high clocked single core cpu of course.

Right now the highest clocked production one is from IBM. The power6.
It runs at 5Ghz.

You realize such a box eats 6+ kilowatt?

By accident the power6 is not the fastest core single core for computerchess. Just accident, that might change with power7 once again.

Now i'm not sure whether you're aware that such machines are pretty expensive. If in 2010 we have again a limit of 8 cores then some $100k+ power7 machine from IBM will be probably fastest. If we allow SMT that is.

Note current rules never disallowed SMT. Intel claims their 'logical cores' as real cores though, weirdly that did get allowed, as Hiarcs box had 'em.

You sure you want next world championship some of the participants show up with a '8 core' machine from IBM that eats a shitload of power?

Happened before with DEC Alpha. In 1997 there was big iron cases cooled to -45C with Kryotech. 633 experimental non-sold alpha's 21164 got overclocked to 767Mhz.

Fastest PC at that time was the P2-233Mhz. Intel was on the brink of releasing the P2-300Mhz. I don't need to mention that such a DEC alpha box, with kryotech service personnel next to the box in the playing hall, is a lot faster than a P2-233Mhz.

In 1999, Chessbase played on Siemens PC's. Hah pc's. You gotta be kidding me. A truck arrived at the playing hall in the morning prior to the matches. We loudly asked ourselves: "what's this?".

While an Siemens official was rolling out of the truck a 500 kilo monster, Mathias Feist arrived. He told us: "that is Fritz".

Junior also had one of course.

So relative spoken, chessbase had the best hardware.

I'm not counting the supercomputers of course, writing software for clusters is not easy huh?

There was 3 supercomputers joining by the way.

From head it was Zugzwang on 512 processor Alpha 450Mhz and P.Conners at a cluster of pentium-2 pc's.

Cilkchess ran on a 512 processor origin3800. I guess that were 500Mhz processors. Cilkchess was a simple bitboard engine and that R14000 was a 64 bits processor.

Funny is that years later i ran also on an origin3800 500Mhz. At 460 processors Diep reached a speed of 9.99 million nps. Average of the last few games (when diep ran smooth and ok, though of course it DID lose 2 of those games) was 5-7 mln nps as minimum each move.

At PC single core, Diep was a lot slower than for example crafty (factor 40).

Pretty amazing i got more nps then huh?

You see, you just don't want to forbid all these guys from showing up.

h) cheating gets harder when opening up hardware; if we play at single cpu or 2 socket hardware, then you better play LOCAL, because if someone plays remote and has at home a 24 core box with rybka 3.0 at it, he can easily cheat on you, as the latest rybka won't be that much better at 8 cores than someone analyzing at 24 cores, so he can knock you out handsdown as he knows which mistakes you might be making.

If you're on a 24 core machine, then the guy cheating already has to use his 24 core box to play in the event.

Note in 2010 a 24 core box is pretty cheap and there will be 48 core AMD machines. End of 2010 intel will show up maybe with a 32 core box (64 logical cores).

It is not about proving someone is cheating, it is about not giving anyone the easy opportunity to profit from it. This went totally wrong in 2009.

Vincent

[Bill Rogers]

Mr Hyatt
I hold great respect for your contributions to computer chess over the years and aldo for your opinions on chess matches.
For the unlimited over all chess champion you are 100% correct but just as the analogy you used about heavy weight fighters in the world of boxing that is why they have different weight classes just as we have different ELO ratings for all chess players.
I don't think that I am rated anywhere near 1,600 elo and the best my little chess program can do is somewhere around that limit too. It is becomming somewhat of a moot question though as all the newer PC's are becomming more powerful every day. My newist laptop is a duo core even though I have no idea on how to benifit from that feature so in the future the computers used in chess tournements should or may not have any real effects. sigh......
Bill
P.S.
I started out with the idea of trying to state that there should be different type of limits on hardware but then it ended up seeming useless.

[bob]

Mr Hyatt
I hold great respect for your contributions to computer chess over the years and aldo for your opinions on chess matches.
For the unlimited over all chess champion you are 100% correct but just as the analogy you used about heavy weight fighters in the world of boxing that is why they have different weight classes just as we have different ELO ratings for all chess players.
I don't think that I am rated anywhere near 1,600 elo and the best my little chess program can do is somewhere around that limit too. It is becomming somewhat of a moot question though as all the newer PC's are becomming more powerful every day. My newist laptop is a duo core even though I have no idea on how to benifit from that feature so in the future the computers used in chess tournements should or may not have any real effects. sigh......
Bill
P.S.
I started out with the idea of trying to state that there should be different type of limits on hardware but then it ended up seeming useless.

There has never been anything inherently wrong with uniform platform chess events. Don Beal used to organize one every year, where all you had to do was send him an executable and a book, he provided the hardware and ran the event. You could attend if you wanted. His idea was to determine which "program" was best. But it was a flawed concept. For example, Cray Blitz depended on the vector hardware of the Cray, which the PCs didn't have, so it ran at a huge speed penalty. Deep Thought could not participate at all since it used special-purpose hardware. Parallel / cluster programs could not participate because only one cpu per program was used. There have been lots of other examples where this was too exclusionary. A program developed for the DEC Alpha could not participate. Programs written in any assembly language but X86 could not participate, even though they would run on processors of an equivalent speed. With all these issues, the event died due to lack of interest by anybody...

We have uniform platform results from CCRL, SSDF, etc. We don't need the WCCC to be a similar format...

[Spock]

We have uniform platform results from CCRL, SSDF, etc. We don't need the WCCC to be a similar format...

Agreed

[Laskos]

We have uniform platform results from CCRL, SSDF, etc. We don't need the WCCC to be a similar format...

Agreed

But uniform book, preparation for each opponent, engine tuning, and so too.

Kai

[sje]

It might be difficult to organize, but how about an event where each machine has an energy usage limit? This could be expressed in watts or in total joules per game.

On my desktop, Symbolic handles about 20,000 nodes per joule; this figure varies a lot. The number would be much higher when if the program were running on a notebook.

[diep]

It might be difficult to organize, but how about an event where each machine has an energy usage limit? This could be expressed in watts or in total joules per game.

On my desktop, Symbolic handles about 20,000 nodes per joule; this figure varies a lot. The number would be much higher when if the program were running on a notebook.

Oh what a ballony, why not limit the number of total nodes searched?

Vincent

[diep]

Mr Hyatt
I hold great respect for your contributions to computer chess over the years and aldo for your opinions on chess matches.
For the unlimited over all chess champion you are 100% correct but just as the analogy you used about heavy weight fighters in the world of boxing that is why they have different weight classes just as we have different ELO ratings for all chess players.
I don't think that I am rated anywhere near 1,600 elo and the best my little chess program can do is somewhere around that limit too. It is becomming somewhat of a moot question though as all the newer PC's are becomming more powerful every day. My newist laptop is a duo core even though I have no idea on how to benifit from that feature so in the future the computers used in chess tournements should or may not have any real effects. sigh......
Bill
P.S.
I started out with the idea of trying to state that there should be different type of limits on hardware but then it ended up seeming useless.

And in the technical sport we are in, did the shrink consider the concept of logical thinking?

Namely that you might end up with like 1 participant?

Initially Hiarcs and Junior join.

Junior wins with fantastic openings play from Hiarcs, just like someone from the Russian chess school would have played it. Hiarcs team protests as they feel junior played like a young god, yet older than junior itself. Hiarcs withdraws leaving 1 participant.

[smcracraft]

It seems to me that limiting any time of human endeavor,
whether tennis, chess, robot-computer-wars, airplane
competitions, boat races, etc. to any multi-class
comparison is absurd.

The *sole* goal should be "best finish" regardless of
resources.

The entirety of the human race has been to pill everything
behind the leader to pull it over the finish line in the best
possible way.

I feel that any holding back from "real excellence" to make
for subcategories, artificially hobbled competition, or
purposely handicapped categories is nuts and completely
contrary to our past as a species.

To coin a phrase of Gordon Gekko's in Wall Street:

Excellence is right.

Excellence works.

Excellence clarifies, cuts through, and captures the essence of the
evolutionary spirit.

Excellence, in all of its forms -- Excellecnce in life, in money, in love, in
knowledge -- has marked the upward surge of mankind.

And excellence -- you mark my words -- will not only drive tournaments
forward, but that other malfunctioning entity called the human chess player...

--Stuart