Chess will never be solved, here's why

Nope, that's just an excuse to never admit you were wrong.  You'll always be able to claim it wasn't enough, or done right, or what have you.  You're a Vickylan in sheep's clothing...

As one (of many) of your opponents, I started your clock when you began playing this game.

When did Sveshnikov say that?

Sadly he died of Covid-19 last August. He was a hero of mine.

#98
Full quote:
"Chess is an exact mathematical problem. The solution comes from two sides: the opening and the endgame. The middlegame does not exist. The middlegame is a well-studied opening. An opening should result in an endgame.... Give me five years, good assistants and modern computers, and I will trace all variations from the opening towards tablebases and 'close' chess. I feel that power."
That was 2 years before he died. I was surprised by his statement. He was right.

#96
No, the number of positions 4*10^37 is not affected by the 50/75 moves rule or the 3/5 fold repetition rule. However without the 50/75 moves rule both sides could play around in a 32-men position for trillions of moves jamming the engine. With neither 50/75 moves rule nor 3/5 fold repetition rule, chess is undecisive. A game 1 Nf3 Nf6 2 Ng1 Ng8 etc. can go on forever.

@tygxc

"No, the number of positions 4*10^37 is not affected by the 50/75 moves rule or the 3/5 fold repetition rule."

Why not?

If the same board layout and side to move can be reached from two tabya you can't assume the evaluation is the same unless the ply counts match when the 50 move rule is in effect. Including the ply count as part of the position multiplies the number of positions by 100. (For the exercise it can be assumed that a losing player will claim under the 50 move rule if possible, so the 75 move rule can be ignored).

idk why yuh guys always believe in AI engines ngl these are dumb and useless

Uhh even if it does ill still hate it xd

#101
You may overvalue the impact of the 50 moves rule on solving chess.  Indeed in the table bases there are positions where a forced checkmate is prevented by the 50 moves rule. That is why in ICCF you can claim a table base win beyond 50 moves. The main impact of the 50 moves rule is to enforce an irreversible move i.e. pawn move or capture. Indeed there is always one player who benefits from claiming the 50 moves draw, i.e. his best move is to claim the draw.
The 5898.5 moves end in a draw with 2 bare kings: all pawns moved in steps of 1 square and promoted and all original and promoted pieces captured.
If I understand correctly your earlier argument goes that there exists a win with 3 men KR vs. K in x moves, a longer win with 4 men, even longer with 5 men, even longer with 6 men, even longer with 7 men, even 400 moves with 8 men.
https://medcraveonline.com/IRATJ/chess-endgame-table-base-records.html#:~:text=Introduction,by%20Lewis%20Stiller%20in%201991. 
https://arves.org/arves/index.php/en/latestnews/latest-news/2-ongecategoriseerd/1509-8-men-tablebase-first-explorations

So you postulate the possible existence of an ultra long win with 32 men. Even if such a 32-men position exists, there is no evidence that it could be forced from the initial position.

This makes no sense to me.

Wins are wins and draws are draws. Humans and computers have no trouble distinguishing what has happened at the end of a game.

I will make one very important point. The percentage of draws does not indicate either player is close to perfect. Karpov and Kasparov had huge numbers of draws but were about 800 points weaker than the strongest current computer (so would have been beaten close to 100% of the time by such an opponent). In hindsight it is clear the results of their matches showed nothing more than that they were both strong and quite evenly matched (with Kasparov being somewhat stronger). Certainly you can't rely on their results as a reliable guide to the true result of chess.

But the same may be true of current top computer games. They are closely matched and very strong, hence they have lots of draws. It may be they are close to perfection, but it is not impossible that there could be another engine that could score 90% against the latest Stockfish.

An interesting exercise would be to see how results between top engines change with alterations to the time allocated. It might be possible to extrapolate to the asymptotic expected score with infinite time. This would indicate where perfection lies on the Elo scale. It is certainly some finite rating, but not known what.

@Elroch

Sorry. Should have been clearer. When I referred to the percentage draw trend, I wasn't referring to the trend over time in practical play, rather the trend over increasing number of men in perfect play (as indicated by the tablebases). The latter trend appears to show that theoretically drawn positions are a feature principally of positions with less than 12 men, the percentage of such positions plummeting rapidly from 100% with 2 men.

There are nevertheless obviously draws with any number of men, but I would expect these to be extremely rare (percentage-wise as the Americans would say) in 32 man positions. Many positions have been evaluated as drawn in opening theory and that works very well in practical play but I would suggest that the discrepancy arises because the people making the evaluations are out of their depth (to mate) in these positions. 

When I say humans see long forced wins as draws, I'm referring to their evaluation of the starting position for the win, not suggesting they don't know when they're mated.

I give additional reasons for the increasing percentage of draws over time in practical play in e.g. #22. There is obviously a difference in practical play between humans and engines, but I think both are so far removed from perfect play that it makes virtually no difference.

Increasing the time allocation for computers may or may not improve their performance in terms of perfect play. You would first of all have to show that it does. See http://izvolitve.ijs.si/Stacks/Articles/19805735.pdf.

Aside from the number of games it would require, has it been shown that the procedure would converge to perfect play?

There would certainly be such obfuscating factors. My whole argument based on published tablebase figures is far from solid. 

A more sophisticated extraction of figures from the tablebases (which are mostly the only thing we know about perfect play) would give a better picture.

But, for the present, I think my view is probably not far from the truth.

#106
"An interesting exercise would be to see how results between top engines change with alterations to the time allocated."
AlphaZero - AlphaZero: 88.2% draws @ 1 s / move, 97.9% draws @ 1 min / move.

#111
How do you explain the many draws in ICCF: they are allowed to claim table base wins beyond 50 moves, but it does not happen, 9 decisive games in 136. So though the >50 moves 7-men wins are there, they are not reached.

@tygxc re #111

The ICCF draws are without exception agreed draws. Not relevant to perfect play in any way. They usually give up well before they reach the tablebase stage.

I already gave an explanation of the phenomenon in #22.

@Elroch gives a less contentious explanation in his post #106.

as long as theres a finite number of chess positions, it is always possible. whether or not its practical is another question.

Does that mean 7 total pieces on the board or 7 pieces per side (14 pieces)?

#113
The position you posted in #22 is a draw. Equal material, symmetrical pawns, neither player can hope to win. How many months should they continue to play that? They do not get tired, they know no time trouble.

The wins >50 moves have in common that they have no mobile pawns. In the initial position there are as many pawns as other men, they are all mobile, they can chose between 1 and 2 steps, and pawn moves are necessary to bring bishops, rooks and the queen into play. Those >50 move wins are an endgame feature without mobile pawns, where the pieces get into long wielded manoeuvres.
https://en.wikipedia.org/wiki/Two_knights_endgame