hgm wrote:I fail to see what "Human error" would add to a computer competition...
it adds the element of the "unknown"
operator ability(or lack thereon) has played a role in this past WCCC and in the WCCC blitz and and in other recent computer events
i recall a recent event in which an operator playing against Rybka knew that Rybka had a bug which would sac pawns to avoid a draw by perpetual and because of this human knowledge the operator ignored all draw offers by Rybka which happily sacked enough pawns until it lost
probably not the exact details here but it was something very similar
i am not looking for the truth here..just an interesting event
Sean Evans wrote:Hi, I think for the 2009 WCCC, the rules should be one CPU of your choice and a set amount of RAM. The 2008 WCCC really doesn't mean much when the programs are running on completely different levels of CPU power!
This discussion pops up every year.
WCCC is "world COMPUTER chess championship". There is no mention of "software" there. You are supposed to bring the best computer contraption you could come-up with. That includes chess chips (Deep Blue), specialized daughter boards (Rebel), FPGAs (Hydra), dedicated chess boards (Novag), PCs, PDAs, etc. It is not about an equal hardware competition.
There is nothing equal about this competition. You bring your best opening book. Maybe even tune it against particular oponents.
It is not about the best program either. Its a combination of software, hardware, books, end-game database, and whatever else that can affect chess play.
We already have tons of same-hardware tournaments (WBEC, CCRL, etc.). There is no need to turn WCCC into that.
Hi lance,
Fully agreed. These discussions are totally unnecessary.
COMPUTER chess is a combination of hardware (the best you can get) and software. They should read the rules at ICGA (imho).
If you want the unknown to play a role, it would be better to have Kriegspiel in stead of Chess.
Of course we could also program the GUI in an automated tournament such that it does not always perform the move the engine comes up with, but sometimes picks another one (e.g. the ponder move of the opponent, or the PV move of a very early iteration of the engin itself). A kind of automated "machine erring". Would that be a good idea?
hgm wrote:
A kind of automated "machine erring". Would that be a good idea?
its a start
but i am afraid you will never be able to teach a program to make human errors
human errors.. thank god.. will never be emulated by the Silicon beast
unless of course you create a program to have the urge to go to the bathroom or feel sleepy or tired and as a result play at elo levels fare below its normal standard
keep in mind that operator input is not only error input but also the knowledge of a particular program(such as weak book lines and bugs) and so forth as i mentioned above
a uniform platform tournament with real red blood ..breathing human operators
at least one a year
now thats REAL computer chess competition regards
Steve
If you want to see something very close to the same hardware look no further than the CCRL, and CETG. there are two organisations that readily come to mind.
Before someone says that they do not use ponder, I already know that!
The SSDF run lots of matches using auto232 and on occasion on different hadware, that is very close to an even contest.
Look at Sadat's site for equal hardware matches.
I would suggest that the WCCC be allowed for entrants to bring a box as big as they can, it is a great advert for the even greater game of chess. For example what if the 64 core machine was used.........against a lowly single core capable program?
I think it would be better to force the engine to send 2 moves and then the GUI would choose which one to play. The first one sent is the one the engine wants to play, and the second the one the engine would play if the first is disallowed. For compliant UCI engines it's as easy as using Multi-PV=2. The GUI could pick a move at random and disable it, if it matches the engine's first move, the GUI plays the 2nd.
Sean Evans wrote:Hi, I think for the 2009 WCCC, the rules should be one CPU of your choice and a set amount of RAM. The 2008 WCCC really doesn't mean much when the programs are running on completely different levels of CPU power!
Sean Evans wrote:Hi, I think for the 2009 WCCC, the rules should be one CPU of your choice and a set amount of RAM. The 2008 WCCC really doesn't mean much when the programs are running on completely different levels of CPU power!
I did not hear anyone complaining when Zapper competed on 512 cores. Also, we all know that to date Zapper scales better than Rybka..I think that a uniform platform tournament would be rather boring, I can do that at my own house.
