I didn't assume "we cannot know that". I just needed more from tygxc than just "I know because everyone knows". He may very well know, and everyone may very well know, but I was looking for some evidence from him, since he was the one making the claim.
Yeah but maybe he doesn't express himself in the way you would prefer him to.
There's no "maybe" about it. He definitely didn't express himself the way I would have preferred.
When you finally realised that you understood one of my posts and that it was correct, and you decided to argue the same way with Elroch, I was hardly holding my breath. ...
I didn't understand the post you're referring to. I try not to understand horsesh*t, otherwise I'd finish up as sane as you.
You had made a statement I happened to agree with, but without any cogent argument in support.
I made, in fact, a very minor point about the shambles of rules laid out in the FIDE laws.
^^ That's for a reason, which it would be impolite to mention,.
Well, I have to say that I'm pleased to see politeness restored somewhere between Posts 5575 and 5577. 
@5561
'Other things being equal, any material gain, no matter how small, means success,' - Capablanca
Steinitz - "I can give God pawn and move and still win"
Since all statements by world chess champions are true (@tygxc assures me) the conclusion is clear.
@5519
"a solution awaits the development of much better engines or a breakthrough in methodology"
++ We already have engines that calculate a billion positions per second.
We already have a methodology.
Start from ICCF drawn games, explore 3 alternative lines at each white move.
Stop calculations when the good assistants determine an obvious draw or loss.
The 10^17 relevant positions can be done in 5 years.
If you deny the good assistants,
then it may well become 5 million years of irrelevant calculations.
Your proposal makes some assumptions I don't accept:
Chess is inherently a draw. Ignoring ICCF won/lost games because "there must have a mistake" could miss a line of play that would indicate chess could be a win.
That the "good assistants" could pinpoint the only three viable alternatives, and thinking there could only be three alternative moves.
Thinking the good assistants will always be correct in their assessment of what will be the "obvious" result.
Believing that present-day machines and programming cannot be relegated to "the dustbin of history" by any revolutionary advances that would make your five million years of calculation a joke.
Again, just because I, or tygxc, or you, or Sveshnikov believes that chess is inherently a draw does NOT make it so. Starting with the idea that chess is a draw and achieving "not a proof or an attempted proof" but simply a reinforcement of our opinions is essentially worthless.
If you had read the post you would realize that it was tygxc that used the phrase 'five million years" to pooh-pooh my belief that a more reliable proof is possible in the long run.
If we are not discussing whether chess can be determined to be inherently won/lost or is drawn, then the OP solved the issue with his observation that as a practical matter chess is already "solved" as a competitive endeavor by the most successful players--what they (man or machine) do is the best chess. He goes on to posit that no algorithms or computer analysis can change that. It would seem that under those criteria the Sveshnikov proposition would do no more to "solve" the game than Capablanca has already achieved.
Therefore the ensuing 5588 entries here are just a waste of time.
Yes and this isn't the only thread there's been on the subject. I'm finding it more interesting than I imagined I would. I think that's due to a bundle of things. I used to be very intersted in computers and progtramming, as a hobby, so that's one thing. I like the combination of deductive and intuitive thought it demands and it's slowly rekindling my old interest. I also think I'm intrigued by the combination of people discussing the subject in this thread, partly because I think there's some dysfunctionality. Normal people probably wouldn't discuss it for the best part of a year, so why? I've also always been interested in writing, itself; and this gives me the opportunity to see how different styles trigger different responses. I still haven't given up the idea of writing a novel, if I'm ever motivated enough to do so. I started one once in the mid 80s and sent it off to a literary agent. She lost it and when I didn't hear anything, I had developed different interests. Then she contacted me telling me she'd lost it and then found it and she wanted me to complete it but by then my life was very different. And these discussions are a welcome distraction from what I've been doing this last couple of weeks, which is sorting out a lot of physical things.
I would call that a "nugget"......
@5585
"Chess is inherently a draw." ++ Yes. There is massive evidence: expert options, ICCF, TCEC, AlphaZero, human games, and the deductive argument that an advantage of 1 tempo < 1 pawn is insufficient to win: you cannot queen a tempo.
"Ignoring ICCF won/lost games because "there must have a mistake" could miss a line of play that would indicate chess could be a win." ++ If you look at decisive ICCF WC Finals games, it is always possible to pinpoint the mistake, usually the last move.
"That the "good assistants" could pinpoint the only three viable alternatives"
++ It is not the good assistants, but the engine that pinpoint the 3 best alternatives once the calculation is running. The good assistants only launch the calculation, i.e. set up the tabiya and they occasionally terminate a calculation e.g. in case of a clear draw.
"thinking there could only be three alternative moves"
++ I previously calculated that, extrapolating from the AlphaZero autoplay paper.
The table base best move is always among the 4 top engine moves if the 10^9 positions /s engine runs for 17 s except for 1 error in 10^20 positions.
You can verify that by setting up a KRPP vs. KRP endgame and letting your desktop run for 4.7 h.
"good assistants will always be correct in their assessment of what will be the "obvious" result."
++ The good assistants only occasionally intervene and only if they are sure.
ICCF (grand)masters do not resign drawn positions or draw in won positions.
"Believing that present-day machines and programming cannot be relegated to "the dustbin of history" by any revolutionary advances"
++ Engines get better. Their error rate goes down. You can see that in TCEC. ICCF WC Finals games now have less errors and thus less decisive games than 10 years ago.
Quantum computers hold the potential to advance in table bases: from 7 to 8, to 9 men...
The idea that you the top three choices of an engine always include a best move is laughable.
Quantum computers hold the potential to advance in table bases: from 7 to 8, to 9 men...
Not for the foreseeable future they don't.
If it's competitive and with humans, no. Why? Because humans aren't computers. People always say "best move", in videos, but not everyone plays the "best move", because a "best move" is only a "best move", if you know that the next move, is also the "best move". I don't ever hear anyone talk about this. Chess engines are very great and good to use and helpful, but they aren't helpful when you don't have them and it's move 36 and you have a clock going down and you've never seen this position before because your opponent is in the same boat as you on move 34. To me this is obvious. I guarantee there are moves that GMs don't understand in certain position that an engine recommends on a given move in a game and it's a game that started off on some "Normal line". Also, another thing that bugs me is that if chess hasn't been 100% been solved, than how can us humans say a move is a blunder when we have to admit we haven't solved chess. Computer engines have recommended moves that we call a blunder, yet chess engines are better than humans. Sooo... Is it really a blunder by definition by the computer? Because back to my first point, you have to KNOW that your opponents next move will be the best one. But the problem is... How can we humans say a move is the best move, when chess engines haven't solved chess? Also, all players, even GMs play different according to their emotions at some point because humans aren't perfect, which makes chess interesting. There is no best moves guys, just like there is no spoon. Anyone else know what I'm talking about? That was a good poop.
@5593
"The idea that you the top three choices of an engine always include a best move"
++ Not the top 3, but the top 4.
Starting from a drawn ICCF WC Finals game, there are 3 alternatives to each move. 3 + 1 = 4.
I arrived at the 4 by extrapolating from the AlphaZero autoplay paper.
https://arxiv.org/abs/2009.04374 Figure 2.
At 1 s / move: 88.2% draws, 11.8% decisive
At 1 min / move: 97.9% draws, 2.1% decisive
Extrapolating:
At 1 h / move: 2.1% * 2.1% / 11.8% = 0.37% decisive
At 60 h / move: 0.37% * 2.1% / 11.8% = 0.067% decisive
60 h / move on the engine of the paper corresponds to 17 s on an engine of a billion positions/s
Assuming game length 37 moves: average of ICCF WC Finals: 1 error in 10^5 positions.
Thus for 4 moves: 1 error in (10^5)^4 = 10^20 positions.
You can verify this.
Take a KRPP vs. KRP position, run it for 17000 s = 4.7 h on a desktop.
Verify the table base exact move is among the top 4 engine moves.
@5596
"how can us humans say a move is a blunder"
++ We have ample evidence that chess is a draw. Thus each decisive game must contain an odd number of mistakes, at least 1. By inspection it is possible to identify at least 1 mistake.
"How can we humans say a move is the best move"
++ We cannot, except for positions with 7 men or less, which have been strongly solved.
We also have statistical evidence. We can say that all the moves of a drawn game in the ICCF world championship finals are > 99% certain to be optimal moves.
For example all moves in this game https://www.iccf.com/game?id=1164259 are > 99% sure to be optimal moves. There is < 1% probability that there are 2 mistakes that undo each other.
Quantum computers hold the potential to advance in table bases: from 7 to 8, to 9 men...
Not for the foreseeable future they don't.
Correct. This is just speculation at the moment. Like pretty much all of @tygxc's "conclusions".
One ridiculous estimate is that only 10^17 positions will need to be examined. Let's consider the last TCEC superfinal games - https://live.chessbase.com/en/Games?id=TCEC-Season-22---Superfinal
The average length of these games (with both sides having free access to a tablebase) is 74 moves. If all analysis games were the same length, a branching factor (adjusted for transposition) of 3 would give 2x10^35 positions. Even a branching factor of 2 would give more than 10^22 positions. As it happens this is a systematic underestimation, because the game lengths go up to 135 moves and the longer games greatly dominate the size of the position tree for reasons that are (hopefully) obvious.
[To illustrate, the average from 1 to 135 of 2^n is 6.45e38]
@5596
"how can us humans say a move is a blunder"
++ We have ample evidence that chess is a draw. Thus each decisive game must contain an odd number of mistakes, at least 1. By inspection it is possible to identify at least 1 mistake.
"How can we humans say a move is the best move"
++ We cannot, except for positions with 7 men or less, which have been strongly solved.
We also have statistical evidence. We can say that all the moves of a drawn game in the ICCF world championship finals are > 99% certain to be optimal moves.
For example all moves in this game https://www.iccf.com/game?id=1164259 are > 99% sure to be optimal moves. There is < 1% probability that there are 2 mistakes that undo each other.
Funniest one yet?
Here @tygxc is justifying one of the steps in his faulty reasoning to show chess is a forced draw by starting by assuming the conclusion that chess is a forced draw! Of course, this is a well-known fallacy.
Classic Fallacy 1: Assuming the Conclusion
It's worth drawing attention to another fallacy implicit in almost all of @tygxc's posts, including the above - arguing that something is probable and then assuming it has become certain. He does this twice in above single post! First he thinks "ample evidence" suffices to prove the conclusion that chess is a draw by induction. No. Not a proof. Then he uses "statistical evidence" to justify a claim he then relies on as certain.
This can never be done in a legitimate proof.
Classic Fallacy 2: Appeal to Probability
@5598
"only 10^17 positions will need to be examined" ++ Yes, that is right.
"the last TCEC superfinal games"
++ TCEC has 50 imposed openings chosen to be slightly unbalanced:
not too balanced to avoid all draws, not too unbalanced to avoid win / loss.
"The average length of these games is 74 moves" ++ They play on too long.
In ICCF the humans would agree on a draw sooner when there is no hope for either side to win.
"If all analysis games were the same length, a branching factor (adjusted for transposition) of 3 would give 2x10^35 positions." ++ That is neglecting transpositions, essential in Chess.
As said before: with width w and depth d
Upper bound without transpositions 1 + w + w² + w³ + ... w^d = (w^(d+1) - 1) / (w - 1)
Lower bound with full transpositions 1 + w/1! + w²/2! + w³/3! + ... = e^w, regardless of d
For an estimate take the geometric mean of the upper bound and the lower bound.
"Even a branching factor of 2 would give more than 10^22 positions."
++ Neglecting transpositions.
"the game lengths go up to 135 moves" ++ Because they play on too long.
"To illustrate, the average from 1 to 135 of 2^n is 6.45e38" ++ neglecting transpositions.
This also proves the upper bound is way too high: there are less than 10^32 positions that can result from optimal play by both players.
https://arxiv.org/abs/2112.09386
@5599
"chess is a forced draw by starting by assuming the conclusion that chess is a forced draw"
++ No. Evidence that Chess is a draw:
1) Expert opinions: Steinitz, Lasker, Capablanca, Fischer, Kasparov, Kramnik, Carlsen.
2) AlphaZero autoplay: more time = more draws, even if stalemate = win
3) TCEC: forced openings to avoid all draws
4) Human top matches and tournaments
5) ICCF WC Finals: g games = d draws + w wins
First assume chess is a forced win. Try to fit a Poisson distribution of errors / game with a probability d / g for an odd number or errors per game. It is impossible.
Consequently assume chess is a draw. Fit a Poisson distribution of errors / game with a probability w / g for an odd number of errors per game.
It is possible and shows the drawn games are > 99% sure to be perfect games with optimal play by both players and < 1% to contain 2 errors that undo each other.
If you really have a degree in statistics, then you should appreciate this argument.
6) A deductive argument. A tempo in the opening is worth 1/3 of a pawn.
You can queen a pawn, but you cannot queen a tempo. A tempo is not enough to win.
First you say an extra bishop does not win and then you say an extra tempo wins.
@5566
I can post you a solution if we accept the definition of solution you are quoting. (I nearly said, "the definition of solution you are using", but of course that would be wrong.) [line again inserted to set context]
"why do you need the supercomputers?"
++ To calculate all 10^17 relevant positions.
That takes 5 years on 3 engines of a billion positions per second.
My point was that according to your definition of "solve" none of that is necessary. Try again.