Page 13 of 16

Re: How about we settle the WCCC argument?

Posted: Thu Jun 21, 2012 7:14 am
by rodolfoleoni
Houdini wrote:1) For example: Houdini, Komodo, Stockfish, Rybka, Critter, Ivanhoe, Fritz, Junior, Shredder, Hiarcs, Chiron, Naum, Hannibal, Spike, Spark.

...............................................................................
Would it be a World Championship, or an exclusive elite tourney?

Re: How about we settle the WCCC argument?

Posted: Thu Jun 21, 2012 7:26 am
by Rebel
hgm wrote:
Houdini wrote:1) For example: Houdini, Komodo, Stockfish, Rybka, Critter, Ivanhoe, Fritz, Junior, Shredder, Hiarcs, Chiron, Naum, Hannibal, Spike, Spark.

2a) A single computer is enough, in which case only a single game is played and broadcast at any time.
2b) Performing "differently on different hardware" is nonsense. The top hardware today is Intel, so you run everything on a good 6-core or 8-core i7.
2c) A program that wants to be "World Champion" and is not MP... joking?

3) Time for a longer match is no problem, as everything is played 24/7 without human intervention. The tournament can take several weeks without any problem, there's no need for the engine authors to be available during the entire tournament. There could be group stages or RR stage, semi-finals and finals.

4) A database of opening positions can be selected by the organizers. For each game there's a random pick, and of course always a second game with reversed colors.

5) The broadcast is 24/7 for several weeks. Engine authors don't need to be available all the time, they can jump in whenever they want.
It sounds like an attempt to maximize the chances of your own engine...

You are weak on books? => Don't allow private books! No cluster version? => Limit the hardware! Only marginally better than the opponents? => Insist on many games!
:lol:

One thing the ICGA was always right is to have 2 titles, one for equal hardware, one for unlimited. Else many programmers won't show up.

Re: How about we settle the WCCC argument?

Posted: Thu Jun 21, 2012 7:31 am
by hgm
Obviously an invitation-only tourney can never be called a 'World Championship'. And if you are excluding the strongest Chess entities in the World by artificial hardware limitations, you could never call it a championship for 'computer Chess'. You might as well require that the program source is not larger than 5KB, to make sure that a certain engine becomes 'World Champion'!

To me it sounds merely like a 'consumer test for middle-game engines'. As there already have been dozens before...

Re: How about we settle the WCCC argument?

Posted: Thu Jun 21, 2012 8:11 am
by rbarreira
hgm wrote:And if you are excluding the strongest Chess entities in the World by artificial hardware limitations, you could never call it a championship for 'computer Chess'.
You could call it a "world championship for PC chess engines".

Re: How about we settle the WCCC argument?

Posted: Thu Jun 21, 2012 10:11 am
by marcelk
Peter Skinner wrote:
Houdini wrote:2) Played on powerful, uniform hardware so that the competition is fair to every engine (no doping!) and doesn't degenerate into a "cluster war".
The problem is there are programs out there that perform differently on different hardware. The only real fair way to play would be on a single core system (Not all programs are MP).
There is more than one way to avoid an escalation into a resource war (if that is really wanted, I don't have a problem with unlimited hardware for a WCCC match).
But if we want limits, instead of specifying CPUs, specify the physical limitations of the system. Eg. by weight or wattage. The everyone can still run on their preferred architecture.

Re: How about we settle the WCCC argument?

Posted: Thu Jun 21, 2012 10:18 am
by marcelk
Houdini wrote: 4) Played with a fixed opening book or set of starting positions selected for the tournament, so that the competition doesn't degenerate in an opening book war. For maximum fairness also play every position with reversed colors.
A lot of development is done my many teams in book learning algorithms and book preparation. IMO booking is an essential part of a WCCC.

If you don't like that, there is Fischer random.

Re: How about we settle the WCCC argument?

Posted: Thu Jun 21, 2012 10:25 am
by Houdini
hgm wrote:Obviously an invitation-only tourney can never be called a 'World Championship'. And if you are excluding the strongest Chess entities in the World by artificial hardware limitations, you could never call it a championship for 'computer Chess'. You might as well require that the program source is not larger than 5KB, to make sure that a certain engine becomes 'World Champion'!

To me it sounds merely like a 'consumer test for middle-game engines'. As there already have been dozens before...
It doesn't need to be an "invitation-only" tournament. Qualifiers or group stages are possible - like in football, Euro 2012.

There is nothing "artificial" about a hardware limitation, you'll find that in most sports. For example Formula 1 limits the hardware to 2400 cc engines: "The engines had to be 90° V8 of 2.4 litres maximum capacity with a 98 mm maximum circular bore, which imply a 39.7 mm minimum stroke. They had to have two circular inlet and exhaust valves per cylinder, be normally aspirated and have a 95 kg (209 lb) minimum weight."
The same in cycling, there are strict limits on the bicycles used.

The goal is the same everywhere: give all participants equal opportunity, and avoid pointless, high-cost hardware wars.

Robert

Re: How about we settle the WCCC argument?

Posted: Thu Jun 21, 2012 10:37 am
by chrisw
Houdini wrote:
hgm wrote:Obviously an invitation-only tourney can never be called a 'World Championship'. And if you are excluding the strongest Chess entities in the World by artificial hardware limitations, you could never call it a championship for 'computer Chess'. You might as well require that the program source is not larger than 5KB, to make sure that a certain engine becomes 'World Champion'!

To me it sounds merely like a 'consumer test for middle-game engines'. As there already have been dozens before...
It doesn't need to be an "invitation-only" tournament. Qualifiers or group stages are possible - like in football, Euro 2012.

There is nothing "artificial" about a hardware limitation, you'll find that in most sports. For example Formula 1 limits the hardware to 2400 cc engines: "The engines had to be 90° V8 of 2.4 litres maximum capacity with a 98 mm maximum circular bore, which imply a 39.7 mm minimum stroke. They had to have two circular inlet and exhaust valves per cylinder, be normally aspirated and have a 95 kg (209 lb) minimum weight."
The same in cycling, there are strict limits on the bicycles used.

The goal is the same everywhere: give all participants equal opportunity, and avoid pointless, high-cost hardware wars.

Robert
yeah, yeah, yeah, everybody believes your reasoning here ;-)

why don't you just suggest to Vas that he enters Rybka in two forms, firstly the giant cluster and secondly on some hardware that roughly parallels yours? The spectators and users ought to be bright enough to work out the consequences of the results ....

Re: How about we settle the WCCC argument?

Posted: Thu Jun 21, 2012 10:41 am
by Houdini
marcelk wrote:A lot of development is done my many teams in book learning algorithms and book preparation. IMO booking is an essential part of a WCCC.
I perceive a lot of dogmatic thinking in the replies to my simple description of how the TCEC tournament became the most successful tournament of the past 2 years.

Of course a lot of work is done by many teams on "book preparation", but is it relevant to decide which engine is best?

As a chess fan I prefer to see some interesting games starting from move 8 or move 10, rather than a book preparation of 25 or 30 moves.
It's funny how HGM used the expression "test for middle-game engines", when in fact this proposal would have engines play the opening instead of relying on 25 moves of theory.

Robert

Re: How about we settle the WCCC argument?

Posted: Thu Jun 21, 2012 10:43 am
by Houdini
chrisw wrote:yeah, yeah, yeah, everybody believes your reasoning here ;-)
OK, I'm out of this thread.
Bye.