A prediction !

[Dann Corbit]

The following long thread in CCC will be about how dishonest was for Rybka to win WCCC 2008 using a cluster with 40 cores !
What do you think? Right or not ?

Regards,
Silvian

Yes, there will always be those who do not understand the rules of the contest and therefore will think something was amiss.

[hgm]

There are hardly any programs that could not run on a PC under Windows. ...

That might or not be true. For example, I have heard of one GPU project going on. That is not going to run on a "standard windows box and hardware".

I would say the fact you heard only of one, while in Chess War nearly 500 programs compete without problems, makes it extremely true. I think that "hardly any" is a quite accurate description of "only one".

And many other past examples, from special-purpose hardware versions (DB, Belle, etc) to specific hardware feature dependencies (Cray Blitz and vectors).

Well, those don't compete now in WCCC either, so in practice it would make no difference. Times have changed in computerland, since Belle.

It is an OK idea, but it is not _the_ way to hold serious tournaments.

This is add total odds with reality. A tournament with only 10 participants, like the current WCCC , is IMO not a serious tournament, especially when you play only one game per pairing. I take a tournament that is a full rund-robin between 60+ participants, like Open War, a lot more serious.

I spent a lot of time doing a parallel search. All that effort becomes wasted effort if run on a single-cpu machine. Someone that did _not_ develop a parallel search therefore has an unfair advantage because they spent more time on the non-parallel stuff.

If you pick the biggest, baddest box available, and let everyone run, that is closer to reasonable. If someone can't use 8 cpus, that's their decision then.
BUt putting everyone into a "one-size-fits-all" box at least stifles development.

I agree that multiple cores is such a common hardware feature now that it should not be qualified as "special hardware requirement". So indeed, it seems reasonable to run each program on 4 cores. Of course, if I were to run such a tournament, for efficiency reasons I would not be so foolish to let engines that can use only a single core waste the power of a quad, but would delegate those to otherwise equivalent single-core machines. (Or run 2 games simultaneously on a dual, if memory bandwidth permits it.)

Another posibility would be to have a eparate tournament for single-core and multi-core engines,just as there is a separate title for blitz and normal time control.

[bob]

There are hardly any programs that could not run on a PC under Windows. ...

That might or not be true. For example, I have heard of one GPU project going on. That is not going to run on a "standard windows box and hardware".

I would say the fact you heard only of one, while in Chess War nearly 500 programs compete without problems, makes it extremely true. I think that "hardly any" is a quite accurate description of "only one".

but "only one" is a stretch of the facts. I mentioned one. I know of others. One that took a very old version of my chess program, moved the eval and generation into some high-powered graphics processor, and played on ICC for a couple of years. There are others. Hydra comes to mind. Uniform platform chess tournaments are about as interesting to most as automobile races where everyone drives the same car...

And many other past examples, from special-purpose hardware versions (DB, Belle, etc) to specific hardware feature dependencies (Cray Blitz and vectors).

Well, those don't compete now in WCCC either, so in practice it would make no difference. Times have changed in computerland, since Belle.

Exactly how? Can you spell H-Y-D-R-A to name one. What about all the parallel search programs, the cluster programs, etc??? Exclude 'em all???

It is an OK idea, but it is not _the_ way to hold serious tournaments.

This is add total odds with reality. A tournament with only 10 participants, like the current WCCC , is IMO not a serious tournament, especially when you play only one game per pairing. I take a tournament that is a full rund-robin between 60+ participants, like Open War, a lot more serious.

And an event like CCTs are more serious than _either_ of those. Authors enter their latest, on the best hardware they can use...

I spent a lot of time doing a parallel search. All that effort becomes wasted effort if run on a single-cpu machine. Someone that did _not_ develop a parallel search therefore has an unfair advantage because they spent more time on the non-parallel stuff.

If you pick the biggest, baddest box available, and let everyone run, that is closer to reasonable. If someone can't use 8 cpus, that's their decision then.
BUt putting everyone into a "one-size-fits-all" box at least stifles development.

I agree that multiple cores is such a common hardware feature now that it should not be qualified as "special hardware requirement". So indeed, it seems reasonable to run each program on 4 cores. Of course, if I were to run such a tournament, for efficiency reasons I would not be so foolish to let engines that can use only a single core waste the power of a quad, but would delegate those to otherwise equivalent single-core machines. (Or run 2 games simultaneously on a dual, if memory bandwidth permits it.)

Another posibility would be to have a eparate tournament for single-core and multi-core engines,just as there is a separate title for blitz and normal time control.

[Tord Romstad]

This is add total odds with reality. A tournament with only 10 participants, like the current WCCC , is IMO not a serious tournament, especially when you play only one game per pairing. I take a tournament that is a full rund-robin between 60+ participants, like Open War, a lot more serious.

And an event like CCTs are more serious than _either_ of those. Authors enter their latest, on the best hardware they can use...

"Serious" is a subjective term. I personally don't regard any computer chess tournaments as serious (although some are more interesting than others; e.g. OpenWar and WCCC are both far more interesting to me than the CCT), and I wish the rest of the community also stopped taking them so seriously. Computer chess is supposed to be fun, not serious. Seriousness leads to competition, secrecy and suspicion, as opposed to cooperation, openness and trust.

It doesn't matter which program wins the WCCC or the CCT, or which program tops the rating lists. What matters is that the general level of all programs keeps improving.

Tord

[hgm]

but "only one" is a stretch of the facts. I mentioned one. I know of others. One that took a very old version of my chess program, moved the eval and generation into some high-powered graphics processor, and played on ICC for a couple of years. There are others. Hydra comes to mind. Uniform platform chess tournaments are about as interesting to most as automobile races where everyone drives the same car...

Exactly how? Can you spell H-Y-D-R-A to name one. What about all the parallel search programs, the cluster programs, etc??? Exclude 'em all???

I don't see H-Y-D-R-A competing in WCCC. So it seems there is no need to exclude it, it is a non-starter. From the current list of participants, it seems only cluster Toga and the mobile phone would have been excluded. I don't think either of them would be a big loss, as Toga could still compete as SMP engine. Especially if you realize that they would very likely be replaced by 20 others, who would no longer be discouraged from articipating by the hardware disadvantage.

And an event like CCTs are more serious than _either_ of those. Authors enter their latest, on the best hardware they can use...

This is a matter of opinion, and I certainly don't share it. I do not consider CCT serious at all, as it is not possible to check if the programs really are what they are supposed to be. I would be child's play to use Toga or Rybka, and disguise its kibitzing with a simple filter so that it can't be recognized, and call it your own engine.

A tournament like Chess War, where each executable can be (and is) scrutinized for not being a cheater, is taken infinitely more serious by me than CCT.

[bob]

but "only one" is a stretch of the facts. I mentioned one. I know of others. One that took a very old version of my chess program, moved the eval and generation into some high-powered graphics processor, and played on ICC for a couple of years. There are others. Hydra comes to mind. Uniform platform chess tournaments are about as interesting to most as automobile races where everyone drives the same car...

Exactly how? Can you spell H-Y-D-R-A to name one. What about all the parallel search programs, the cluster programs, etc??? Exclude 'em all???

I don't see H-Y-D-R-A competing in WCCC. So it seems there is no need to exclude it, it is a non-starter. From the current list of participants, it seems only cluster Toga and the mobile phone would have been excluded. I don't think either of them would be a big loss, as Toga could still compete as SMP engine. Especially if you realize that they would very likely be replaced by 20 others, who would no longer be discouraged from articipating by the hardware disadvantage.

And an event like CCTs are more serious than _either_ of those. Authors enter their latest, on the best hardware they can use...

This is a matter of opinion, and I certainly don't share it. I do not consider CCT serious at all, as it is not possible to check if the programs really are what they are supposed to be. I would be child's play to use Toga or Rybka, and disguise its kibitzing with a simple filter so that it can't be recognized, and call it your own engine.

A tournament like Chess War, where each executable can be (and is) scrutinized for not being a cheater, is taken infinitely more serious by me than CCT.

I'm not a fan of _any_ event where author participation is not possible. That's been one of the best ways to discover new ideas over the past 40 years...

[Tord Romstad]

A tournament like Chess War, where each executable can be (and is) scrutinized for not being a cheater, is taken infinitely more serious by me than CCT.

I'm not a fan of _any_ event where author participation is not possible. That's been one of the best ways to discover new ideas over the past 40 years...

I don't understand your point. From this point of view, ChessWar and CCT are roughly equal. In both, author participation is possible, but not required. However, ChessWar is much more pleasant for those of us with very limited time for computer chess, because it is possible for us to come and leave precisely when we want. In my experience, the discussions during ChessWar also have a higher signal/noise ratio, and are a more effective way to discover new ideas (at least it was like that the last time I checked, but I haven't been logged on to either event any time recently).

Also, from the perspective of exchanging ideas, over-the-board events like the WCCC are of course vastly superior to any online tournament. Nothing beats meeting the other programmers in person.

Tord

[hgm]

I'm not a fan of _any_ event where author participation is not possible. That's been one of the best ways to discover new ideas over the past 40 years...

This makes it obvious you don't have the slightest idea what Chess War is. I visit there every day, to chat with other authors.

I agree with Tord that nothing beat meeting other authors face to face. But that doesn't make an over-the-board tournement automatically serious Chess events. They are principally social events. It is not of the slightest importance to win, only to go there.