@Ethan_Brollier you'll notice how tygxc provides no evidence for his claims, and how his logic is verbatim "they drew so they are perfect".
you'll also notice that the 10^17 does not match the increased complexity whatsoever.
in fact, this 10^17 assumes literal perfect gameplay WITHOUT LOOKING AT ANY OTHER POSITIONS, among many other falsehoods.
@Ethan_Brollier you'll notice how tygxc provides no evidence for his claims, and how his logic is verbatim "they drew so they are perfect" . . .
Yeah, there were a few times in the other topic I thought I was exaggerating his stance, but then he'd reply by agreeing...
How do you know they played perfectly?
Because the game was a draw.
How do you know the theoretical result of chess?
Because those perfect players played it.
-__-
Oh, so now you don't think a perfect game is a draw?
Thinking it's a draw and knowing it's a draw are different things.
No difference. I think things, and therefore they're proven. I do this all the time.
I don't think you know what "proof" means.
Yes I do, for example I know some ICCF games were perfect...
[Go back to top of dialogue]
and pointing this out is truly " personally [attacking] a valued member of the community baselessly" (no shade to you Ethan but you seriously should have trusted me instead of giving tygxc the benefit of the doubt, I'm still waiting on you to change your initial post to reflect that, so as to not mislead other visitors)
@140
"Games relying crucially on an engine"
++ No, the games rely neither on the engines, nor on the human jockeying them.
A remarkably strange claim.
Just to be clear, what exactly do you think they used to choose the moves then? Tea leaves?
Oh, I forgot one minor thing when pointing out how spectacularly far from a proof-tree ICCF chess is. The two players in a game do not even share their analysis (apart from the moves played on the board). Thus is it actually 232 even more abject failures to solve chess rather than a single abject failure.
humans: make errors. engines: make errors. tygxc: since it is humans and engines, it cant make errors!
@146
"what exactly do you think they used to choose the moves"
++ It does not matter how they selected the moves, only the moves themselves matter.
The end result is a draw. That justifies all black moves as fit to draw.
It puts to question all white moves as unfit to win.
So the only question left is if there are some reasonable alternative white moves.
Chess can be solved. If you were to hypotheticaly See all possible outcomes and timelines- For example Super advanced AI or some hypothetical HIgher dimensional Allien being or Gods or whatever. Then Chess is already solved.
Did you even take into account the amount of possibilities of chess games? This is what you're missing, janko
To be fair, he was referring to the fact that it could hypothetically be achieved, not that it is practical. Purely mathematically the answer to the question of "can chess can be solved?" is "yes", just like the answer to the question "can you factorise a random billion digit integer?". In both cases there is an algorithm guaranteed to achieve the objective. In both cases, the algorithm is impractical to execute.
@149
"Did you even take into account the amount of possibilities of chess games?"
++ The number of games does not matter, only the number of positions.
10^44 legal positions
10^37 legal positions possible from 1 box of 32 chess men.
10^38 legal position possible from 1 luxury box of 34 chess men with spare queens.
Of these 10^17 relevant to weakly solving chess.
@149
"Did you even take into account the amount of possibilities of chess games?"
++ The number of games does not matter, only the number of positions.
10^44 legal positions
10^37 legal positions possible from 1 box of 32 chess men.
10^38 legal position possible from 1 luxury box of 34 chess men with spare queens.
Of these 10^17 relevant to weakly solving chess.
reminder that tygxc is making several logical errors here. there is no justification for the 10^34, and calling the 10^17 "relevant" when the 10^17 (by his faulty calculations) would literally be the solution tree itself, and other positions are necessary (and therefore relevant) in order to figure out the tree itself.
humans: make errors. engines: make errors. tygxc: since it is humans and engines, it cant make errors!
I also anticipated that tygxc would try to so ordain.
Trying to argue that the addition of a GM on each side would mean tygxc is 'right' because he says so.
But that last part is nothing new from T.
@146
"what exactly do you think they used to choose the moves"
++ It does not matter how they selected the moves, only the moves themselves matter.
The end result is a draw. That justifies all black moves as fit to draw.
It puts to question all white moves as unfit to win.
So? The same thing happens to us all when we fail to win.
So the only question left is if there are some reasonable alternative white moves.
And then you surprise me by getting it partially right.
White had approximately 2000 chances to vary in those games, most leading to more possibilities that it is possible to analyse. If you are want to include black's moves in a proof tree, you first of all need to deal with all of those "alternative white moves". Then those generated 1 move further down the new lines, and so on until there are no loose ends. Perhaps 10^30 nodes (can't be precise, but certainly an impractical number).
@161
"White had approximately 2000 chances to vary in those games" ++ So you mean 100 games * 20 chances to vary per game? Average game length was 39 moves. With 19 theory moves following a well trodden path from data bases, 39 - 19 = 20 seems reasonable indeed.
"more possibilities that it is possible to analyse" ++ That is what the 17 ICCF finalists did,
that is what they used their 2 servers of 90 million positions/second for.
"If you are want to include black's moves in a proof tree" ++ No.
As the games ended in draws there is no need to question black's moves, only white's moves.
@161
"White had approximately 2000 chances to vary in those games" ++ So you mean 100 games * 20 chances to vary per game? Average game length was 39 moves. With 19 theory moves following a well trodden path from data bases, 39 - 19 = 20 seems reasonable indeed.
"more possibilities that it is possible to analyse" ++ That is what the 17 ICCF finalists did,
that is what they used their 2 servers of 90 million positions/second for.
"If you are want to include black's moves in a proof tree" ++ No.
As the games ended in draws there is no need to question black's moves, only white's moves.
That's actually the reason that they aren't good evidence for chess being a draw ..... because the paths are so well-trodden. It's a shame that the team isn't thinking well.
That;s actually the reason that they aren't good evidence for chess being a draw ..... because the paths are so well-trodden. It's a shame that the team isn't thinking well.
Back to reality...the ICCF is not a team trying to solve chess. They are just playing a tournament, one where the human players pretend they are having an impact, but are just running multiple engines and then selecting engine moves. You could write an algorithm and some kind of wrapper application that would do essentially same thing...run half a dozen different engines and select a consensus engine move using a set of criteria and weightings. The difference is, such an app would be more consistent and score better than the human "jockeys", if those jockeys were to rely on their own intuition or expertise.
That;s actually the reason that they aren't good evidence for chess being a draw ..... because the paths are so well-trodden. It's a shame that the team isn't thinking well.
Back to reality...the ICCF is not a team trying to solve chess. They are just playing a tournament, one where the human players pretend they are having an impact, but are just running multiple engines and then selecting engine moves. You could write an algorithm and some kind of wrapper application that would do essentially same thing...run half a dozen different engines and select a consensus engine move using a set of criteria and weightings. The difference is, such an app would be more consistent and score better than the human "jockeys", if those jockeys were to rely on their own intuition or expertise.
tygxc is presenting it as "trying to solve chess".
Are they Correspondence players or otb players? We have the British Correspondence Champion in our chess club. His otb rating is around 1450 FIDE. What I mean is, is that he wouldn't present good guidance for an engine.
tygxc is presenting it as "trying to solve chess".
Are they Correspondence players or otb players? We have the British Correspondence Champion in our chess club. His otb rating is around 1450 FIDE. What I mean is, is that he wouldn't present good guidance for an engine.
...which is like saying that Formula One drivers are a team working to improve combustion engine technology. The fact that you have to say "presenting" tells the story, so kudos at least for not completely believing Tygxc.
tygxc is presenting it as "trying to solve chess".
Are they Correspondence players or otb players? We have the British Correspondence Champion in our chess club. His otb rating is around 1450 FIDE. What I mean is, is that he wouldn't present good guidance for an engine.
...which is like saying that Formula One drivers are a team working to improve combustion engine technology. The fact that you have to say "presenting" tells the story, so kudos at least for not completely believing Tygxc.
Not sure which part of reality you inhabit but kudos for at least being in touch with a small part of it.
@161
"White had approximately 2000 chances to vary in those games" ++ So you mean 100 games * 20 chances to vary per game? Average game length was 39 moves. With 19 theory moves following a well trodden path from data bases, 39 - 19 = 20 seems reasonable indeed.
"more possibilities that it is possible to analyse" ++ That is what the 17 ICCF finalists did,
that is what they used their 2 servers of 90 million positions/second for.
No they didn't. They just played chess in an unco-ordinated way. As good as a bunch of blind people trying to water an entire planet by pissing in the dark for five minutes without any discussion or map.
By your figures, if they were co-ordinated and did what they needed to instead of playing chess against each other, they might manage 1 part in 10^12 of the job.
"If you are want to include black's moves in a proof tree" ++ No,
No proof tree = no weak solution
The reason is that the definition says:
So you need to recursively prove the result for all legal play by an opponent.
As the games ended in draws there is no need to question black's moves, only white's moves.
Firstly, you haven't proven the result of chess is a draw. Secondly, if you had, you still need to "question" both sides moves - you've just claimed the value of the game is the same for both.
Very odd that you think one of them gets a pass because white scores 55% in non-optimal chess.
the starting position is moving closer to 0.0 than an advantage for white when the engines update, meaning that as engines are improving, they are seeing chess as a forced draw more and more
But that's with engines at the same current top level playing against each other.
It wouldn't come out that way if they were playing other engines of ten years ago.
Engines continue to improve steadily.
Engines of now could draw engines of ten years in the future?
Some of the time. Maybe.
There's information about the most modern engines drawing each other.
But how much info about the other thing?
---------------
What happens to the old software?
Do they store it in a vault somewhere?
Would programmers want it to become very public that the old software 'loses'?
Think about it.
@140
"Games relying crucially on an engine"
++ No, the games rely neither on the engines, nor on the human jockeying them.
They start from the initial position and after the strongest chess on the planet in average 39 moves they end in a certain draw.
That in retrospect justifies all black moves as fit to draw.
You can only question if the reasonable white moves are exhausted.
"blunders in 6 and 7 piece table base positions"
++ Won positions are irrelevant to weakly solving Chess.
Stockfish blitz games are irrelevant: ICCF is average 5 days/move.
ICCF WC Finals are stronger than Stockfish:
that is why finalists got through Preliminaries, Semifinals, Candidates, Finals.
The human is important, that is why 4 of 17 finalists are Russian,
despite worse hardware because of sanctions.
"chess is about 10^24 times as complex as checkers"
++ No. In terms of positions 10^38 / 5*10^20 = 2*10^17 times more.
In terms of required effort to weakly solve 10^17 / 10^14 = 1000 times more.