Chess will never be solved, here's why

#2214

"Thus your sample involves uncertain conclusions based on a sample comprising about 1 of each 10^36 positions. It is virtually worthless for drawing conclusions."
++ That is right: I try to draw conclusions about 10^36 legal and sensible positions from a sample games data base of 10^7, i.e. 10^8 positions. However, the mathematician Tromp is allowed to draw conclusions about the legality of 10^44 positions from a sample of 538 legal positions.

"and the vast majority of these cannot be reached in a reasonable game with > 50% accuracy.
You know you are guessing wildly. So do we all."
++ It is an educated guess. Pick any one of the 538 legal Tromp positions and try to construct a game of > 50% accuracy that reaches it. You are sure to fail. I gave one example of a Tromp legal position and its shortest proof game above. Even more: Pick any of the 10,000 positions without any promotions that Tromp randomly sampled and try to construct a game of > 50% accuracy that reaches it. You are sure to fail again. I gave one example of such a position and somebody else was so kind to provide a proof game. Its accuracy is near 0%. Tromp himself conjectured that 1 in 10^6 positions without promotions to pieces not previously captured may be sensible. Multiplying the number of positions without promotions to pieces not previously captured by 10, to account for positions with 3 or 4 queens leaves 10^38 positions. The conjecture about 1 in 10^6 positions then gives a range of 10^38 - 10^32 legal positions that can occur in a game with > 50% accuracy.

"Some of ICCF or even human games may already be ideal games with optimal moves.
So what? Why would you even think this is relevant?"
++ This is relevant, because it means some heuristics derived from human play, ICCF play and TCEC play may be used to simplify solving chess.

Whenever a pawn promotes, it is to a queen, unless it must be a knight to promote with check or to deliver a smothered checkmate, or unless it must be a rook or even a bishop to prevent stalemate. That is something we know. The solution to solving chess is not suddenly going to include 9 underpromotions.

The center is more important than the wings. It is something we know. That is why if 1 e4 and 1 d4 are proven draws, then it is not necessary to check 1 a4. It could be done on demand, but there is no point in wasting resources on it as the outcome is beyond any doubt.

Losing a pawn or an exchange without clear compensation usually leads to a loss. Losing a piece even more. That is something we know. Thus it is pointless to investigate 1 e4 e5 2 Ba6. We know it loses for white.

#2255
"I guess you forgot that Pfren, BlueEmu, and Llama...all of whom are rated significantly higher than you"
++ Please kindly remind me what Pfren, BlueEmu, and Llama wrote on this thread here.
I have trouble finding any of their statements in the 2258 posts. Maybe you can help me.
Pfren, BlueEmu, and Llama are different persons from me, so it is only logical that they have sometimes different opinions than me on some topics and sometimes the same. The whole point of a discussion forum is to confront different opinions.

Tromp waved you off when you tried to push your BS...10^44 dropped to 10^36 dropped to 10^17...and I am guessing he did so for the same reason we are.  It doesn't matter if only 1 in a million positions turns out to be viable/worth evaluating...because you have to evaluate them anyway at some level to make that determination which will take longer than any of our lifetimes.  You just keep pretending this will magically happen with no effort whatsoever.

I'm not going to search it all...but Llamas posts were fairly recent, so I'm pretty sure you can find them.  Are you planning to claim that these posters do *not* disagree with your numbers?  Because most of the regulars on these type of threads here know that they did/do.  The point is, you reduced the set of posters down in a dubious fashion that only serves your interests...which is *exactly* what you did to come up with your 5 year plan .

Oh, that's the point? Nice that you didn't write the names in reverse order. I wonder what is Schaeffer's rating in checkers, though...

And to distinguish proofs from hypotheses, whenever possible. But of course it is proven that intelligence, education and logical thinking skills are proportional to chess rating, isn't it? So none here can be better than you at that...

#2259
"Tromp waved you off when you tried to push your BS...10^44 dropped to 10^36"
++ No, Tromp could not agree with any of my proposed definitions of a sensible position.
My present best effort: sensible position = legal position with a proof game of > 50% accuracy.
Tromp was quite helpful generating the random sample of 10,000 positions without promotions to pieces not previously captured. Tromp conjectured that only 1 in 10^6 of those positions without promotions to pieces not previously captured might occur in a reasonable game.

"It doesn't matter if only 1 in a million positions turns out to be viable/worth evaluating...because you have to evaluate them anyway at some level to make that determination"
++ No, I do not have to evaluate them, as they do not occur during weakly solving chess.
Only positions that are legal, sensible, reachable, and relevant need evaluating as they turn up during weakly solving chess.
Illegal positions do not turn up as they cannot be reached from the initial position.
Non-sensible positions like with 2 white dark square bishops or 5 black rooks do not turn up as they cannot be reached from the initial position by a game with > 50% accuracy and we are searching for an ideal game with optimal moves and thus an accuracy near 100%.
Non-reachable positions do not turn up. Many positions with a white pawn on e2 are legal and sensible, but none can be reached after 1 e4.
Non-relevant positions are legal, sensible, and reachable, but do not matter. If 1 e4 e5 is proven a draw, then it is not relevant if 1 e4 c5 draws as well or not.

#2255
"You are also using rapid ratings of human beings, and pretending they should have some meaningful impact on understanding how chess can be solved by engines or some future technology."
++ I did not pretend or even imply that, that is your interpretation. Maybe the ratings have some meaningful impact on how humans behave on the forum. How much ad hominem arguments they use. How much sarcasm. How much ridicule. How much they gang up.
Just an example: "Your rating gives you not one ounce more credibility than you have ever had here, which is not much."
In your own style: "Your low rating gives you not one ounce more credibility than you have ever had here, which is not much."

#2261
"And to distinguish proofs from hypotheses, whenever possible." ++ Yes, that is all right.

"But of course it is proven that intelligence, education and logical thinking skills are proportional to chess rating, isn't it?"
++ I did not write or even imply that.
This forum is open to all. Rating, university degrees, age do not matter.
Btickler mentioned a relationship between certain people. I hinted at a possible relationship.

Interesting.

There are many ways to offend. Mixing up proofs and personal hypotheses as if they were the same thing, for example, can be viewed as a form of manipulation; citing a paper to support one's claim, when the paper in fact does not, is another. @tygxc, e.g., repeated dozens times a wrong citation about the number of nodes searched to solve checkers, but only after 7(!) requests to pinpoint it in the paper, he (or she idk) conceded it was wrong. If he understands as he says, it's not possible he didn't understand that point, and in fact I think he understood perfectly; but that's only worse than not understanding.

As for understanding each other, I don't think we cannot understand you two because of a lack of intelligence or education, and there is no conspiracy or a "gang" against you. It's because your definitions and the relationships between them are not clear at all, compared to those one can find in textbooks and papers, and that one knows s/he understand well from discussion with peers, grades, exercises, etc.

As for the relationship between ratings and the capability to understand the topic, imho there is basically none. I hope you don't want to suggest that the higher the rating, the more intelligent the person. One can think that there is a relationship only if s/he thinks that her supposedly better evaluation function (yes, humans use an evaluation function too, but the rating does not depend only on that) can be used to understand how to reduce the search space, in order to weakly solve chess. I reject this hypothesis entirely, because evaluation functions are based on inductions from a limited number of situations and therefore can be affected by faulty generalization. A single "black swan" can subvert an expected score and that's why basically nobody proposes to cut down branches from the search space as @tygxc proposes to do. As for this point:

In fact, they mention it all the time. They call it "broad search", for example, and they clearly distinguish it from the weak solution.

So you've reached the 7 man tablebase in this drawn position.

And you've done all the takebacks up to the first position and find they're still drawn.

What's SF14 going to suggest as your first alternative move from the first position? (You can look it up if you like.)

(As an aside - strictly irrelevant - any progress on my challenge to prove just a single one of Tromp's samples can't be reached by play with an accuracy > 50%?)

#2267

"There are many ways to offend." ++ Sure there are. Like you did.

"citing a paper to support one's claim, when the paper in fact does not, is another."
++ There were two different papers by Schaeffer about solving checkers. One mentioned 10^9 nodes, the other 10^7 nodes each needing 10^7 positions. Hence the confusion.

"I don't think we cannot understand you because of a lack of intelligence or education"
++ It is about ability or will to understand. I do not care about intelligence or education. I try to explain in a way that is understandable to all. I answer all questions to the best of my ability.

"there is no conspiracy or a "gang" against you"
++ Most of my arguments get just met by some insults, some ridicule, some sarcasm, some unsound counterargument, plain denial, some trolling. That then gets applauded by another one. I have no problem at all with different opinions, as long as it stays civil and constructive.

"It's because your definitions and the relationships between them are not clear at all"
++ I try to define as clearly as possible. I try to indicate the relationships as clearly as I can. I am aware that my reasoning may sometimes be too succinct for some to understand, but I do not want to make my posts too long either.

"As for the relationship between ratings and the topic, imho there is basically none."
++ Basically little. A little bit maybe: there is or rather was a relationship between chess playing strength and education in science. Anderssen, Steinitz, Lasker, Euwe were mathematicians. Capablanca, Botvinnik, Vidmar were engineers. That however is not what I implied. Rather like: we cannot beat this guy in chess and neither in science, but surely we are better at insulting, ridiculising...

"their supposedly better evaluation function"
++ I do not believe in an evaluation function to solve chess, only in deep calculation and using heuristics derived from human and engine play. The evaluation function will always fall short. Only deep calculation until the endgame table base or an earlier 3-fold repetition of position is reliable.

Any chance of stopping bickering and answering some of the outstanding questions?

Does your reference to 3-fold repetition above mean that you've decided on the game you want to solve or are you reserving the option of switching to 2-fold repetition should it prove convenient?

@tygxc is not 'trying to offend'   (unlike the other guy)
but there's his pattern of not letting anybody tell him anything -
combined with the immortal and extremely revealing "know more than any of you".
He is not trying to be offensive per se.  Not his goal apparently.
But it ends up amounting to the same thing.  
He's doing better than the other guy also - in the sense that he doesn't have the obsessions with chess ratings and his own rating that the other constantly manifests.

As to what @tygxc 's goal is here or if there is one ... perhaps its to do with 'proof of solution'.
'Solved'.  And 'proof'.  Which has legitimate connection to the forum topic.
I keep wanting to give him benefit of 'doubt' about Everything -
because he does try to be civil.  ('pretty sure I know more than any of you' ...  is there a 'paradox' there ?)   

#2268
Here is again the shortest proof game for the first legal Tromp sample.

Its accuracy is near 0%. You can add some moves and/or change the move order, but that changes nothing: the accuracy stays near 0%.

As for the other position it is obvious that white must promote his pawn: preferably to a queen, or else to a bishop, or to a rook, or to a knight. What does that prove? Of course promotion to a queen is the move.

#2270
"Does your reference to 3-fold repetition above mean that you've decided on the game you want to solve or are you reserving the option of switching to 2-fold repetition should it prove convenient?"
++ It does not matter. For simplicity of argument keep it 3-fold as it is written in the Laws of Chess. In practice I would use 2-fold to save the pointless extra repetition. If 2-fold repetition is the optimal for white and for black, then so is 3-fold, and 4-fold, and 5-fold. The result would be the same even if the Laws of Chess were ammended to 7-fold repetition. 3-fold repetition is just a practical rule, that allows to save some time on the clock. For example in this game black did repeat 2-fold to reach the time control at move 40.
https://www.chessgames.com/perl/chessgame?gid=1084375 

You do have an appallingly slapdash approach.

In the game you mention Black won with the 3-fold repetition rule in force. With a 2-fold repetition rule the game would have been drawn.

Is that what you mean by "The result would be the same"?

But can we now pin you down on the game you propose to solve as basic rules + 3-fold repetition rule. If your computation has to check with you what rules it's supposed to use before each move, you're never going to reach 10^9 nodes a second, even on a cloud cuckoo machine.

#2271

"there's his pattern of not letting anybody tell him anything"
++ Oh yes, for example I changed my phrase 'excess promotions' to the longer 'promotions to pieces not previously captured' because of your remark. I also multiplied the 10^37 by 10 to include positions with 3 or 4 queens. I also formulated more precise definitions of some terms.

'the immortal and extremely revealing "know more than any of you".'
++ It is not something I would normally do. I do not boast about academic titles, careers, publications, ratings... but I got provoked by being accused of not knowing mathematics. At the same time somebody with 2 degrees or somebody working at an university gets qualified as 'very knowledgeable'. I am pretty sure I know more about mathematics than any here on this forum thread. That then gets ridiculed as if I qualified myself as 'the greatest living mathematician' which surely I am not.

"As to what @tygxc 's goal is here"
++ I have only one goal: to discuss weakly solving chess and GM Sveshnikov's claim.
GM Sveshnikov was the visionary who made his bold prediction.
I only and modestly try to verify it and I still believe he was right indeed.

#2274
"In the game you mention Black won with the 3-fold repetition rule in force. With a 2-fold repetition rule the game would have been drawn."

++ No: if the rule were 2-fold, then black would not have repeated the moves, or he might have lost on time. If the rule were 5-fold repetition, then black might have repeated the moves 4 times, to gain time on the clock and check and double check his calculation.

There it is yet again:
" I am pretty sure I know more about mathematics than any here on this forum thread. "