Chess will never be solved, here's why

@7030

"At least not in the near future"
++ Depending on somebody willing to fund a 5 year project of 3 million $ with 3 cloud engines and 3 grandmasters.

I costed it a couple of hundred posts back. You can no longer fund even a 5 year project with $3m, sorry. I speculated that if I got someone like my son to lead it, he probably wouldn't do it for less than £140K per year and the reason for that is there wouldn't be much hope of it leading anywhere, so in effect it would be five wasted years at the most productive point in someone's career. It may be that anyone who would do it for less wouldn't be competent anyhow because the decision to do it may not be a very competent one.

Anyway, you seem to think that the methodology would be known. If fact it would take around five years just to plan the five year project, which would turn into a much longer one, of course. You seem to imagine that if you started now, the team would automatically be competent, the premises required would be available and the team would know exactly what hardware and software they required. In fact, the first part of the project would be getting the hardware and software together. The project you imagine would last five years at $3 million is really 10 years and $10 million and that would be the minimum requirement to get it off the ground. Of course, it would still be a fiddlingly small project, inadequately supported and with no real chance of success.

Incidentally, 10 or 15 years ago I designed the overview of a real experiment on the paranormal which had adequate checks and balances. The hard part of that is to prevent the team of controllers from wrecking the project. That would take, I thought, three years of training: getting people with disparate views working together, before the experiments could begin. Two years of training and planning at a minimum. The cost would be phenomenal to do a proper series of experiments on the paranormal. I was getting irritated by all the incompetent university departments who were believing, in the 90s and 00s, that they had "debunked" the paranormal where in actuality they had designed experiments that prevented positive results from being found.

Whether or not you believe in paranormal possibilities, I think that if you're intelligent, you'd at least accept that you have to listen to people who may genuinely understand something about it, because if you don't, clearly your experiments are worthless.

From a slightly different angle, which is not all that different, a five year project to weakly solve chess is also worthless. Who would get involved, knowing that the overview was so badly contrived? Not thought out .... contrived.

Oh so I did guess right that it was a win for black? I thought it was a black win just by looking at it for a few minutes, as you will remember. Like you, I didn't think any human could find it over the board. Maybe a year with an analysis board? My reasoning was that on an open board the two bishops should actually win, from that position. So it was just a guess, although an educated one.

@7034

"Why does stockfish need to reach the table base." ++ Because the evaluation of the 7-men endgame table base is exact draw/win/loss and the evaluation of Stockfish is only approximate.

"You have already claimed Stockfish plays perfect chess."
++ Not my claim. My claim is that at 17 s/move on a billion nodes per second engine the optimal move is always among the top 4 engine moves with 1 error in 10^20 positions.

"Can anyone tell me why Stockfish can not solve this position"
++ Stockfish can solve this position if you give it enough time to reach sufficient depth.

"And when you run Stockfish you will see Stockfish is hitting the table bases"
++ Once the 7-men endgame table base is hit play is perfect. 

That seems to be a REALLY good argument against ty's perpetual nonsense.

Yet you claim that Stockfish is capable of actually solving chess, which means completely accurate assessment of positions. Dishonest or forgetful?

@7039

This position is not relevant and cannot result from the initial position by best play from both sides.

For "geologically slow", you seem to not have gotten the context, someone suggested infinite speed, which obviously means nothing will be slow.

As for your bold text, it also misses the mark. The idea is, suppose you got a mate in 5 (M5) position, all opponent moves will lead to lines that are mate in five. Maybe it is easier to assume that the winning side is also forced to mate in 5 moves first, to get the point across. SF will easily find this in almost no time, and it does so by brute force testing all lines. Normally when SF does a tree search, it does not get to a mate in most leaf nodes of the search tree, and it has to use its strategy function to estimate the value of a position. But this is not the case when there is a mate or a direct draw (like stalemate) on the board. For this case, it does not have to make this estimate/guess, it will just slab on its score for win/draw/loss. If all leafnodes in its search tree are either mate or direct draw, there is zero guessing, all it has done is brute force checked every line to mate/draw. Now, you can do the same with M6, same result. But if you try something like M30, it at best will take quite some time to find this forced mate. All I am suggesting is that no matter how high you set the value n for mate in n, it will do the same thing, and find a perfect answer with no guessing. And since the maximum number of moves in a legal game is about 5000 moves, (a quick google suggested 5899, I think this is inaccurate, but it is something like that), by setting your search depth at that number, you should be able to force the engine to go this deep in every line, again giving zero guessing, only a brute force search.

Someone suggested that SF does not work exactly like that, I have not bothered to look into that. I do know that simple Alpha Beta searches work the way I described, I have not studied the latest tree search tricks, but it would surprise me quite a bit if you cannot make the engine brute force to a certain search depth quite trivially.

Edit: For last paragraph, after a quick google:

https://www.chessjournal.com/stockfish/#Does_Stockfish_use_Minimax_algorithm

In other words, it appears that Stockfish does not use the Shannon Type B search algorithm, but rather a variant of Minimax, meaning that my first paragraphs are correct, and the someone who suggested SF does not work like that, is mistaken.

@7042

"the maximum number of moves in a legal game is about 5000 moves, (a quick google suggested 5899" ++ 5898.5 due to the 50-moves rule. https://wismuth.com/chess/longest-game.html However, from the initial position a 7-men endgame table base draw or a prior 3-fold repetition is reached long before 300 moves.

Because it's a win for black, of course. But the assessment is not really about that position. It's about stockfish and your ridiculous claim that with a little guidance from 3 grandmasters smoking cigarettes in long holders and sipping gin and tonics, SF is sufficiently accurate to guide chess to a weak solution ..... completely accurately.

It seems that if Mr DisparateKingToddle is to be believed ... and I see no good reason to disbelieve himtockfish is incapable of performing at a level of accuracy which you have previously claimed is possible and which we know would be necessary. 

Just out of interest - can the initial position be reached from the initial position by best play from both sides? 

If not, are there any positions that can be reached from the initial position by best play from both sides?

Do tell us!

@7045

"can the initial position be reached from the initial position by best play from both sides?"
++ The initial position is the initial position.

"are there any positions that can be reached from the initial position by best play from both sides?" ++ Yes, more than 10^17 and less than 10^44.
Example
https://www.iccf.com/game?id=1164344

In 57 moves from the initial position to a 7-men endgame table base draw, from one drawn position to the other.

Oh thanks! Now we know.

P.P.P.S. "Candour"

Chandelier.

At least he talks to you. He completely stopped acknowledging my existence a year ago, when I made it clear I know that he isn't very honest.

Maybe using chess engines is unreliable, a programme that stores every possible move in a huge database and detects which sequence of moves leads to a checkmate might answer our question. But then, we don't know which is a forced win...