Mike S. wrote:Unlike during the 1980s/90s, when there were two tournaments separated from each other, maybe world champion titles in two categories could be awarded in one event, alltogether.
(1.) The absolute Computer Chess World Champion
(everything unlimited)
(2.) The Uniform Platform Computer Chess World Champion
(
consumer-orientated hardware limit; other limits like for opening books are debateable IMO but I won't stress that.)
Some will say, what do we need (2) for, or I am only interested in (1). Others will say, (1) is meaningless for me, or with the equipment I can afford I have no chance in (1). So,
this proposal is only for people who can accept, and want to handle compromises. A compromise can provide the chance to include all parties of interest, unlike when an extreme position wins and others are excluded (for whatever reason or in whatever role).
I think it can be done if up to two entries are allowed, per participant. But in that case, one entry would have to play for (2.) with limited hardware. In case of only one entry, he should have the free choice of category. Maybe a long swiss tournament could be done in a way that whenever possible, the category of the opponent is switched, round by round (like the switch of colors).
The idea of uniform platform raises the problem - which maybe also is a chance though - to find a sponsor for the required number of identical computers, on site.
Alternatively, the particpants would need to bring in "very similar" hardware for category (2). But I think that is only the second best choice, because even with adjustments, they can probably not be standardized in the sense of uniform platform. - Or the requirement for / definition of "consumer-orientated hardware limit" is regulated in a way that it allows a certain bandwidth. Limiting the number of cores only, seems very general.
Also, for now - but probably not for the near future - I think 4 cores would be a better limit, because as explained by Jens, there are huge cost and performance differences within an 8 core limit, still. We have had similar thoughts here, because the same topic was being discussed at the Rybka forum some time ago too, where I had suggested u.p. with quads.
As for the verification of how many cores are used: This just cannot be a problem in a place where computer experts gather together. For example, Microsoft's Process Explorer doesn't even need to be installed, so the tournament director could have it on an USB stick. So nobody can hack it to display a faked CPU load (in case if lovers of conspiracy theories had such concerns
). There are also free CPU identification tools, like CPU-Z.
http://technet.microsoft.com/en-us/sysi ... 96653.aspx
(I am almost sure that such ready-to-run tools also exist for Linux and Mac, but I don't know it.)
Anyway, I am aware myself that my proposal is not very realistic, but it should show that a compromise is possible, at least as something "theoretically imaginable"...
This "compromise" is basically identical to just event 1 existing, because no one will care about event 2, unless it has a very attractive prize (which I doubt will happen).
We already know how various engines stand with "consumer-oriented hardware limit", there is enough data accumulated and published online.
"Consumer-oriented" event may at first sound like a good idea, but to just limit hardware is only one step towards the consumer. For an event to be really consumer-oriented, only publicly available engines should be allowed to compete. With publicly available books as well, and on available common hardware.
Then of course there is question how do the consumers actually use a chess engine. If most of the consumers use chess engines for analysis, rather than for playing with them (which I believe to be true), then using a shared shallow opening book would give a better picture about how well various engines are suited to this purpose.
Then there is a question: what conclusion can a consumer draw from a result of such an event? (event 2). Obviously the statistical significance is too low for the result to be of any use for consumer's decision-making. Although world title may be useful for marketing, it is actually misleading (as most people don't know how to interpret tournament results), and therefore is against the interst of the consumer.
So the whole idea of the consumer-oriented event is flawed from about every angle.
Plus, with ongoing testing projects, providing thousands of games under a variety of testing conditions, the consumer already knows all he needs to know about the comparative strength of chess programs.
What consumer in fact wants from a WCCC is the battle of the most advanced computer chess players that could be created. Consumer wants to be thrilled. Uniform Platform even would not thrill anyone. No one will even watch it, other than true fanatics like us.
Clearly the "Uniform Platform", or "Limited Platform" is not interesting for the consumer. As evident from this and other forums, it is also not interesting for the absolute majority of programmers. The natural question is then: who needs a hardware limit?". Probably only ICGA needs it, lobbied by those commercial programmers (are there more than 1?) who already possess an 8-core hardware.
The reason why those commercial programmer(s) would want such limit is not only to weaken the competitors (eg Rybka). I think they foresee that with hardware limit there simply will be much less of competition at all. Because many authors will decide to ignore the event. So those commercial programmers have double benefit. They will play on their fine-tuned 8-core machine, with a few remaining opponents. And with a good probability they will then display "World Champion 2009" in their ads.
Question is: How the ICGA could fall for this, and if there are other factors (unknown to us) that influenced their decision.
Note that form the ICGA point of view the decision is totally stupid, as it destroys the world title, and without the world title, who needs the ICGA? May be ICGA will have more success organizing Hex of Connect 6 tournaments?
Best,
Kirill
