#141
At least it is a scientific approach. It even corresponds with the human knowledge derived from millions of human games. The engine reinvented the Berlin, the Grünfeld etc. with no external input but the Laws of Chess by playing 700,000 games against itself.
Here is another AlphaZero paper on opening theory
https://arxiv.org/pdf/2111.09259.pdf
Chess will never be solved, here's why
Sort:
Jan 23, 2022
0
#124
#141
It's essentially going through the same process that humans have used over the centuries, so you'd expect it to come up with much the same openings.
#143
You may well believe 1 Nh3 wins in 800 moves or even 1 e4 b5 wins for black in 700 moves, but there is no evidence at all to support that.
The consensus among experts is that chess is a draw and the results of classical, ICCF and TCEC games corroborates that.
With or without 50 moves rule should make no difference.
Bigger rules changes make no difference either.
Jan 23, 2022
0
#126
Experts are expert at playing at a human level. There's absolutely no evidence that bears any relation to perfect play. (But lots of evidence that it doesn't in the perfect play we know about i.e. tablebased endgames - much simpler than the opening.)
Without some way of relating the two I'd discount any evidence based on practical play. The ICCF and TCEC games corroborate nothing.
I don't believe, "1 Nh3 wins in 800 moves or even 1 e4 b5 wins for black in 700 moves", I've no idea where the flaws in opening theory are, but at a reasonable guess I'd say the there are many. I think the fact that nobody mentions the 50 move rule in opening books, but the 50 move rule according to the perfect information we do have appears to assume greater importance as the number of men increases, is probably an indication of just how far we are removed from perfection.
Jan 23, 2022
0
#127
Optimissed wrote:
Perfection in chess is any move that doesn't lose in a hitherto perfectly played game.
@Optimissed #146
A perfect move depends only on the position from which the move is made, irrespective of whether previous play is perfect. If that position is a theoretical win in the game in question, it would be a move that doesn't draw or lose.
Whether a chess move is perfect depends on which kind of chess, basic rules or competition rules.
In the latter case the ply count has to to be taken as an attribute of the position.
So in this position
White to play, ply count = 0
White may make 49 moves between b8 and c8 and each move would be perfect (as would many others) but from the position reached only Rc1# would be a perfect move under competition rules. No other move would lose, but any other move would draw. Under basic rules any move (including Rc1#) would be a perfect move.
Under competition rules perfect play by one player is play in which all moves by that player are perfect.
Under basic rules perfect play is play in which all his moves are perfect and for each position occurring that is theoretically winning for him there is a natural number n (possibly 0) such that mate is played after n moves. An infinite series of perfect moves is not perfect play under basic rules.
Jan 23, 2022
0
#128
@Optimissed #147
I think your proof needs a little working on.
I gave this hint before, but 'll repeat it.
If you purport to prove something about the starting position and your proof contains nothing that distinguishes the starting position from some other position, then, if the proof is valid, what is proved must also be true of the other position. If it's not true of the other position that means your proof is missing something.
In this case your "proof" contains nothing intelligible that distinguishes the starting position from say this position.
White to move, ply count = 0
I'm still working on what the rest says, but you can probably tell me. Does what you purport to prove apply to that position?
Jan 24, 2022
0
#129
@Optimissed #151
The quote you give was part of a comment on #146 viz.
"Perfection in chess is any move that doesn't lose in a hitherto perfectly played game."
which involves the implicit assumption that chess is a draw.
If by "your situation" you mean the diagram I posted in #148 and by "my proposal" you mean the italicised phrase above, then obviously I can't apply your proposal to my situation because, with your assumption that chess is a draw, my situation could not arise with a "hitherto perfect game". If the situation did arise then, in the line I suggest, Black is forced to make the moves Ka1 and Kb1 both of which lose so it would be impossible for him to continue the game with perfect moves according to your definition.
How does your proposal change under competition rules? Under those rules perfect play (presumably what you mean by "perfection in chess") corresponds simply with making perfect moves. It's under basic rules that a caveat is necessary.
Jan 24, 2022
0
#130
Optimissed wrote:
MARattigan wrote:
@Optimissed #147
I think your proof needs a little working on.
I gave this hint before, but 'll repeat it.
If you purport to prove something about the starting position and your proof contains nothing that distinguishes the starting position from some other position, then, if the proof is valid, what is proved must also be true of the other position. If it's not true of the other position that means your proof is missing something.
In this case your "proof" contains nothing intelligible that distinguishes the starting position from say this position.
White to move, ply count = 0
I'm still working on what the rest says, but you can probably tell me. Does what you purport to prove apply to that position?
What proof are you talking about? I know you're confused even if some others can't tell. Firstly, my opinion is my opionion, which I think has a much better chance of being correct that yours does on this subject and secondly, if you demand proofs of someone, even though they weren't offered, it would be good if you could base your demands on a necessity which can be proven accurate. As it is, you're making all sorts of claims and that's all they are.
In #147 you say, "It's logically impossible for there to be a forced win for black, since ...".
I assumed the strange scribings that followed were intended as a proof (though I couldn't discern anything I'd usually call a proof).
If that wasn't the case I withdraw my suggestion that your proof might need some working on (though your English might).
#155
Checkers was popular and interesting too and was not 'designed' to be solved, and yet it has been solved.
Jan 24, 2022
0
#133
it was solved
you can even stalemate your opponent to make them lose
Jan 24, 2022
0
#134
The claim "Opening play largely stays the same even with more radical rules changes." is not very well supported by the paper. There are large changes in the evaluation of lines described in the paper.
#159
See Figure 4: largely the same results across the rules variants.
Also "Stalemate=win chess has little effect on the opening and middlegame play" - Kramnik
Even without adding new (different) pieces, the board size can be increased to 10x10 or FR could become more popular. I like increasing the board size. Kramnik has a good idea in eliminating castling in one move.
uhhhh
Jan 24, 2022
0
#140
crazyhouse & bughouse are my favorites... 960, capture anything, 3-check & 4-player chess are all cool too & i like the 'fog of war' concept, but 'no-castling' just seems to kill off the ability to spice the game up with opposite-side castling & i can't really get into the idea. (nice pfp DH) 😆
#138
"Perfect opening theory is probably very different depending on whether it's for basic rules or competition rules"
Probably not at all.
Opening play largely stays the same even with more radical rules changes.
https://arxiv.org/pdf/2009.04374.pdf
Well that's back to using AZ to verify.
As I said playing Nalimov is sufficient in practical play. E.g. to win "frustrated wins" against the SFs under competition rules. But not against Syzygy.
KNNKP mates can be at most 128 moves deep. You wouldn't expect AZ to give a correct answer on perfect play in the opening where there are likely to be mates a trillion or more deep. You would expect it to give a correct answer for its own level which would be good enough for practical purposes.