World Championship Winning Computer Chess Software Program & Downloads for Chess Databases, Analysis and Play on PC, Mac, iPad and iPhone — Visit: Hiarcs.com
Gian-Carlo Pascutto wrote:
Sooner or later, the Rybka team will have to suck it down too. It looks like it will be later, but the day will come.
I think that the important question is whether computer chess will be an interesting field anymore when that happens. My guess is that it will not--my guess is that when this happens, some program will have an elo of 3400 and Rybka will not ever be able to beat it because the 100 point difference due to having an elo of 3500 will only be useful against "weak" opponents who are "only" around 3000-3200.
I think you misunderstand how the ELO system works.
An 100 ELO advantage is just as big if you are 2000 rated than if you are 5000 rated.
Gian-Carlo Pascutto wrote:
Sooner or later, the Rybka team will have to suck it down too. It looks like it will be later, but the day will come.
I think that the important question is whether computer chess will be an interesting field anymore when that happens. My guess is that it will not--my guess is that when this happens, some program will have an elo of 3400 and Rybka will not ever be able to beat it because the 100 point difference due to having an elo of 3500 will only be useful against "weak" opponents who are "only" around 3000-3200.
I think you misunderstand how the ELO system works.
An 100 ELO advantage is just as big if you are 2000 rated than if you are 5000 rated.
The numbers are relative, not absolute.
I know how the Elo system works--my point was that when you're at an elo of 3500, you're so close to perfection that the differences come about not by how you perform against relatively similarly-rated opponents, where all games are drawn, but by how well you thrash lower-rated opponents compared with someone who is relatively close to your strength. In other words, it is no longer a Gaussian distribution.
turbojuice1122 wrote:
I know how the Elo system works--my point was that when you're at an elo of 3500, you're so close to perfection that the differences come about not by how you perform against relatively similarly-rated opponents, where all games are drawn,
I do not believe there's any evidence that this would be the case ("all games are drawn").
ricard60 wrote:Chess is not solved and is not going to be solved in the next 300 years until then we do not know how a high elo will behave.
ultra high elo regards
Ricardo
I could not disagree more. From all practical purposes chess will be "solved" within the next 10-20 years. Top computer games will essentially all envolve around fighting for advantages of a few centi pawns and no game will in practice get even close to a win. And I would expect this to happen even if one side (white or Black) is having far less resources (e.g. time, processors etc) than the other side.
In fact just a few years from now I would expect that ALL leading engines will have a solid setting that when combined with a huge opening book always hold the game for white. Maybe 15 years down the line the same is possible for black.
What might not happen the next 300 years is that we have a "proof" that chess is a draw in the same sense as checkers has been proven to be a draw. However, the fact we cannot "prove" its a draw has little to do with what can be established in practice.
There is no doubt that white objectively speaking has a won position if black plays without his queen. Yet, we might not have "proved" this in 300 years.
The number of different items (games) that can be disputed in a board is 10 raised up to the power 123, that is to say, a 1 followed (consecutive) by 123 zeros. This number is major that that of atoms that exists in the universe. It wants to say that it does not exist no computer will exist in the world that could calculate so many plays ever. Only the quantum computers were coming closer…
But possible games do not matter as they will contain positions already solved in other games, so what matters is numbers of positions, that's about 1*10^43.
We will never manage to discover the perfection in the chess, we will not also know if the white ones have advantage on the negresses, but it that you say is certain Dylan, but only with the near(next) generation of computers when we happen (pass) from the atomic computers that are the current ones to the quantum computers.
Into the near (next) future there was changing the measurement ELO due to the fact that there was coming a moment that was stagnating and no machine was advancing anything, already this Rybka giving the first step and there is no any more remedy that to extend the Elo up to the " 4000 Elo "
IA wrote:The number of different items (games) that can be disputed in a board is 10 raised up to the power 123, that is to say, a 1 followed (consecutive) by 123 zeros. This number is major that that of atoms that exists in the universe. It wants to say that it does not exist no computer will exist in the world that could calculate so many plays ever. Only the quantum computers were coming closer…
Regards....
It is interesting that this flawed argument still is repeated ad infinitum!
The first problem is that the notion of "proof" as "mathematical proof" in some sense is too strict. From a purely mathematical perspective we might not be able to prove that white has a won position if black plays without his queen and two rooks.
The second problem is that virtually all mathematical arguments do much better that just exhaustive consideration of all possibilities.
There are many ways in which its mathematical possible to prove a statement without exhaustive search. There is for example a simple mathematical argument that Hex is won for white on a board of size n x n for any number n.
The third problem is that game can be established as drawn for reasons that are intuitive clear (when combined with computer analysis). If all top computer games always have evaluations of plus/minus a few centi-pawns even if one engine is running on a (future) mobile phone, while the other engine is running on 1.000.000 cores on sensible person would suggest the game is anything but a draw.
