sachin884 wrote:
this forum is going till 5k posts which consists nothing but trash
Sorry you had a (probably involuntary) break in "screen time", but don't take it out on others... 
Chess will never be solved, here's why
Sort:
Thats absurd comes out of the condition of the game or checkmate. If we dont accept the draw as a solution in chess. Some french champions deluded by their opening are trying to make legal the stalemate confito
trying to make legal the stalemate condition as a winning end. That resolvef all abdurds of chess engines mess.
Aug 21, 2022
0
#3305
sachin884 wrote:
this forum is going till 5k posts which consists nothing but trash
I bet it doesn't stop then.
Aug 21, 2022
0
#3306
tygxc wrote:
@4108
"what happens when an imperfect player evaluates a position?"
++ Neither the player nor the engine evaluates the position,
the 7-men endgame table base evaluates the position and it is perfect.
You appear entirely delusional here. The large majority of the positions in those games have more than 8 pieces on the board. You claim "perfect" analysis of those positions. This includes (hilariously), the claim that every single opening played is known to be perfect.
These positions CANNOT BE FULLY ANALYSED to a tablebase. Doing the much simpler task for checkers required a great deal of computation. Doing this task for chess cannot be done at present.
"This happens in every decisive game." ++ Yes, every decisive game has some error.
These are decisive games played against the engines you are relying on for 100% reliable evaluations.
"All available sources of opinions are imperfect and can be completely wrong."
++ Yes that is true, but some absolute statements are sure not to be wrong.
That is as useful as the trading advice that some currencies are sure to go up. The question is which ones.
Certain endgames with opposite colored bishops are drawn.
True, but not relevant.
1 a4 cannot be a better move than 1 e4 or 1 d4
Your blunder (which I am qualified to point out) is in the jump from good bet to certainty.
Your reasoning is exactly the same as buying a lottery ticket and saying "I am certain this ticket will not win".
You are right to be confident. You are wrong to be certain.
1 e4 e5 2 Ba6 loses for white
A more extreme example of the same.
"It's a good bet that on an individual position, a very strong player is correct in their evaluation."
++ I even calculated that probability.
No, you did not. You estimated it from a sample. An estimate from a sample is NEVER 100% reliable. That is a statistical fact.
If you let the 10^9 nodes/s cloud engine run for 17 s, or a desktop for 4.7 hours, then the absolutely correct move will be among the top 4 engine moves in all but 1 case in 10^20.
I am not 100% sure if this is a wrong estimate or an uncertain estimate, but it is certainly not a hard fact.
sachin884 wrote:
this forum is going till 5k posts which consists nothing but trash
And you’re back? 🤡
Aug 21, 2022
0
#3308
Optimissed wrote:
You are right to be confident. You are wrong to be certain.
1 e4 e5 2 Ba6 loses for white
A more extreme example of the same.
Elroch, this is where you fall flat. Sticking to this principle of "we cannot know anything" is as wrong in principle as it's wrong in practice.
There is no excuse for believing I believe that. I have often directly contradicted it.
White giving away a bishop on the second move of a game loses by force and we can know that.
On the contrary, we know all facts that have been proved. One of those is that checkers is a draw with best play.
One could equally criticise you for your own certainty, after all. You condemn yourself, except in the eyes of a nihilist. Your own insistence that tygxc is wrong fails to your own principles. It's too much a mixed message.
It is true I have absolute faith in the rules of logic.
I also have great respect for the entirely different inductive reasoning of science, which leads to confidence but never absolute certainty.
But unlike some people such as tygxc , I never confuse the two.
To return to original post, what does "Chess being solved" even mean?
Is it finding the best move for every possible position?
Aug 21, 2022
0
#3310
@Ralphmcm, there are a few distinct types of solution. Here it has been generally acknowledged while we would be interested find an "ultra-weak solution" (proving what the result of chess with perfect play is, but failing to exhibit strategies to achieve it) but that it is extremely unlikely that a useful one exists, and have focussed almost entirely on the notion of a "weak solution". This consists of a complete, practical strategy for each player that guarantees getting the theoretical result.
Part of this (and in practice achieved simultaneously, as for checkers) is to prove what the result of chess is with best play by both sides.
ok?
Aug 21, 2022
0
#3311
Optimissed wrote:
Elroch wrote:
@Ralphmcm, there are a few distinct types of solution. Here it has been generally acknowledged while we would be interested find an "ultra-weak solution" (proving what the result of chess with perfect play is, but failing to exhibit strategies to achieve it) but that it is extremely unlikely that a useful one exists, and have focussed almost entirely on the notion of a "weak solution". This consists of a complete, practical strategy for each player that guarantees getting the theoretical result.
Part of this (and in practice achieved simultaneously, as for checkers) is to prove what the result of chess is with best play by both sides.
ok?
You really ought to study my posts more. Thee only viable strategy lies in "playing the best moves". These definitions have been written by idiots. They have no significance for the solution of chess.
While I have read your posts, this article and its ilk is what merits study. Your point does not have substance.
Do you really have the narcissistic arrogance to say the authors are "idiots"? They understand that what the work discussed has achieved is a weak solution. I am also 100% sure they understand ultraweak solutions of other games and why the distinction matters. They also very clearly understand why a weak solution is very different to a complete tablebase for checkers (which would provide a strong solution).
It's always the other guy in Oblivimissed-world .
Optimissed wrote:
Not your argument, btickler and also, when either you or Elroch are concerned, that's very true. Statistically, that is.
Thought you said you didn't want me to troll you and yet here you are, trolling. No doubt you'll say I started it. Obviously, what Elroch said to me wasn't starting it, was it.
I'm not arguing any points in discussion. Just making a rather apparent observation. Toughen up, it hardly rises to the low water mark of your usual fare.
@4107
"The large majority of the positions in those games have more than 8 pieces on the board."
++ The whole point is to calculate until the 7-men endgame table base,
or a forced 3-fold repetition, or a sure draw like some opposite colored bishop ending.
"the claim that every single opening played is known to be perfect"
++ Each opening is known to be perfect after chess is weakly solved:
after the calculation shows that black can draw against all reasonable white moves.
"Doing this task for chess cannot be done at present."
++ 10^17 positions is 1000 x more than 10^14, so chess is 1000 x more complicated, but now can be done in 5 years. It is not because chess is more complicated that it cannot be done.
Chess is even 100 million times more complicated than Losing Chess, despite the same 64 squares and the same 32 men.
"These are decisive games played against the engines you are relying on"
++ No, I do not rely on any decisive games, I rely on drawn games only.
'for 100% reliable evaluations" ++ The evaluation comes from the table base or from a prior 3-fold repetition or a prior human adjudication of e.g. an opposite colored bishop ending.
"the jump from good bet to certainty."
++ That is no bet, it is certainty. Once 1 e4 and 1 d4 are proven draws,
then it is absolutely 100% certain, no doubt at all, that 1 a4 cannot win for white either.
Likewise we do not know, if 1 Nf3 is any better or worse than 1 e4 or 1 d4, but we know for sure that once 1 Nf3 is proven to draw, that then 1 Nh3 cannot win for white either.
"1 e4 e5 2 Ba6 loses for white A more extreme example of the same."
++ It is absolutely 100% certain, no doubt at all that 1 e4 e5 2 Ba6 loses for white.
I even proved it is a forced checkmate in 82 @3936.
"You estimated it from a sample."
++ Yes, I estimated the 1 error in 10^20 positions from a sample.
There is some margin for error. Maybe it is not 1 in 10^20 but 1 in 10^19 or 1 in 10^22,
but that does not change anything fundamental. It is still less than 1 error in 10^17 positions.
"If you let the 10^9 nodes/s cloud engine run for 17 s, or a desktop for 4.7 hours, then the absolutely correct move will be among the top 4 engine moves in all but 1 case in 10^20.
I am not 100% sure if this is a wrong estimate or an uncertain estimate"
++ Yes, this is an estimate: it may well be 1 case in 10^19 or 10^22.
You can see for yourself. Set up any KRPP vs. KRP, let your desktop run for 4.7 hours, verify that the table base exact move is among the top 4 engine moves as predicted.
Aug 22, 2022
0
#3316
tygxc wrote:
@4107
"The large majority of the positions in those games have more than 8 pieces on the board."
++ The whole point is to calculate until the 7-men endgame table base,
or a forced 3-fold repetition, or a sure draw like some opposite colored bishop ending.
It is ridiculous to suggest that the opening position can be exhaustively analysed to a puny 7 piece tablebase. Chess has over 10^29 times as many positions as are in this tablebase. That's over 100,000,000,000,000,000,000,000,000,000 times as many.
The solution of checkers used a tablebase that was relatively a lot larger (in terms of log complexity)
"These are decisive games played against the engines you are relying on"
++ No, I do not rely on any decisive games, I rely on drawn games only.
Perhaps you are unaware, chess engines play chess against each other, and occasionally they lose. Glibly saying a little more time will make them unbeatable is not good enough for several reasons obvious to several of us.
Aug 22, 2022
0
#3317
Optimissed wrote:
Ralphmcm wrote:
To return to original post, what does "Chess being solved" even mean?
Is it finding the best move for every possible position?
Finding the best moves in reasonable positions, since there's no real point in looking at random positions or those where one player has played so badly that they're obviously losing. For instance, perhaps by having played 1. e4 e5 2. Ba6, which Elroch seems to think may not be losing for white. Seems that tygxc and he make a wonderful pair. So similar in so many ways.
I am probably more confident that white is lost after 1. e4 e5 2. Ba6 than that white cannot force a win from the standard starting position. I also recognise that this is an unproven result, according to the standards of mathematics, game theory and computer science, which correctly don't view results as definitely true based on inductive reasoning.
Optimissed wrote:
No, I'm going the other way and becoming less tolerant of bad manners by others. If you want me to behave nicely after a narcissist calls me a narcissist because he can't actually think of a refutation of my post (because there isn't one) well I jolly well shan't, so there!I think that a lot of people here think that game theory consists of the strategy of games. It doesn't.
As if you were ever remotely tolerant .
I don't see anyone conflating game theory with solving chess here, except you. Do you also conflate game theory and gamification?
@4130
"the opening position can be exhaustively analysed to a puny 7 piece tablebase"
++ That is also the expert opinion of the late GM Sveshnikov.
"Soon all openings will be given exact scores, and if the variant is correct, it will lead to a technical endgame in which a draw will be achieved with accurate defense."
"The solution of checkers used a tablebase that was relatively a lot larger"
++ The size of the tablebase does not matter much. log complexity is highest at 26 men and then decreases. See table 3. https://arxiv.org/pdf/2112.09386.pdf
"chess engines play chess against each other, and occasionally they lose"
++ They lose very rarely if given enough time and if no unbalanced openings are imposed like now necessary in TCEC to prevent all draws.
They lose when they make a mistake, that is their top 1 move is not the right move. If they can always take back up to 3 times per position they never lose.
@4131
"don't view results as definitely true based on inductive reasoning"
++ Apart from expert opinions and inductive reasoning based on AlphaZero, TCEC, ICCF, there is also deductive reasoning.
1 pawn is enough to win: by queening it.
1 tempo is not enough to win: you cannot queen a tempo.
1 bishop is enough to win: trade it for a pawn.
I even provided proof that 1 e4 e5 2 Ba6 is a forced checkmate in 82. @3936
this forum is going till 5k posts which consists nothing but trash