Chess will never be solved, here's why

The demented parrot strikes again! 

I'll repeat what I wrote in post #7503:

++ I have even quantified the error rate: 1 error in 10^5 positions for a 10^9 nodes/s engine calculating 17 s/move.

The flaws in your method have already been pointed out.

You say here 

Your desktop is 1000 times slower than a cloud engine of 10^9 nodes/s. Time * 60 gives 5.6 times less error. 

If you were to look at these games as you steadfastly refuse to do, you will notice that four of them were played at 37 mins. per move. According to your figures, 17 sec. per move on your cloud engine is equivalent to about seven and a half times the time I  allocated on my desktop, so according to your "calculation" these games should have 1 half point blunder in around 42,500 ply.

The four games have a total of 290 ply so according to your "calculation", the expected total number of half point blunders in the games is about 0.007.

User @cobra91 has carefully checked the actual total with the Syzygy tablebase here. It comes to 11.

YOUR CALCULATIONS DON'T WORK. CAN YOU STOP POSTING THEM, PLEASE?

The answer is obviously ++No.

Except that in this case any person's willingness to "take a risk" with their conclusion is, in fact, wrong, by definition.  This topic is not about whether chess is a draw.  It's about whether it can be proven to be a draw.  The answer is no, for the foreseeable future of humanity.

Scroll up to the top of page.  Now read the thread title.  Do you see the word "draw"?

Where did Euler say that?

Few rational people would.

Dumb question I suppose. Let’s say chess is solved by a super computer. Does it change anything for us? We can’t even figure out how to beat a sub optimal stockfish.

No, you just have to use words in a different way from everybody else, as you claim the right to do. In this case you use the word "know" to refer to the concept that others refer to as "guess".

Not sure what you use "guess* for, given that you say  you know and it isn't a guess. Probably if you have paranormal abilities you can mean the same and different things in one sentence. 

That would depend on whether the solution is a relatively short win or not. 

If it turns out to be forced win in say less than a hundred moves it would be possible to follow and we may as well all give up playing.

If it turns out to be a draw or a win in 2000 moves it would make very little difference,. Youl'd just get different opening theory with a lot more lines, but it would be impossible to follow for any great number of moves, 

     True, but if the question is "Can we prove that chess is a win/draw?" then not being able to disprove either question is no guarantee of the truth of its opposite. 

     Believe it or not, there are many people who don't take the consensus of expert opinion or the evaluations of top computers to be the gospel truth and continue to hold their own opinions.

    Should it be proved to be a win in 70 moves (or whatever) this should change nothing for human players. Suppose that 1.e4  c5 is found to be black's best choice to delay loss as long as possible. White would not be able to just memorize the best line all the way to checkmate, but know EVERY possibility of a second or third best move by black on move 11, or 18, or 25, or 30, and so on. There may be a handful of people on earth who could perform this miraculous feat of memorization, but it would not help them should black play 1.....e5, or e6 or c6 or a6, or d5, or g6, or d6. For 99.99999% of chess players the solution would have no practical value.

     This doesn't mean we wouldn't all like to know the answer.

I think you're probably right. Hard to say.

If it's a mate in 70 it's likely to be a pretty specific line for each response and the variations will funnel into a limited number of continuations. You then wouldn't need to remember all the responses, just a few "stepping stone" positions along the routes, nor would you need to remember responses that take you into easy mates.

But perhaps I should rather have said mate in 12.

     Of course if there were a forced win in twelve moves top-flight chess would by finished, but this seems extremely unlikely. In the case of longer wins, there would have to be so many lines stemming from every black (assuming it is solved as a win for white) first move, and every black second move from each of those moves and so on that the task of memorization would be untenable for all but top GM strength players with phenomenal memories, even under the criteria you posit.

     This would not apply should a series of moves be discovered that will win no matter what the opposition plays, but it also seems highly unlikely that such a system would exist. Different responses must be made against different inferior moves if the winning positions are to invariably be reached.

With a team of 17 grandmasters and one supercomputer we can prove there is no win in 12 moves.

You only need a big red telephone for that.

This position was argued over for a century, half the time it was believed to be a White win and half the time a draw,

The longest possible distance to mate with a pair of blocked pawns on the g file and a White pawn on the h file is 33 moves.

So if there were a mate in 70 from this position

 it would already have been found.

Really?

There are almost more errors in your logic then lines of text in your comment

there is no justification for the decisive game/ time per move function u use.

you falsely assume 1 error per decisive game

you falsely assume some sort of comparison between alphazero and your '10^9 nodes' thing

you ignore the fact that there were pre set starting openings.

you falsely assume that your program could catch any errors made.  

you falsely assume that different nodes are of comparable strength.

you falsely assume that the 'top 4' are going to have equivalent individual chances of finding the correct moves.

you also assume a pre existing set of positions to evaluate.  that set doesnt exist yet.

your jump to the "10^9 nodes" is so baseless that it is hard to even begin with what goes wrong with it, because there isnt any reason given for it in the first place.

my best guess at what you erred there was to conflate an alphazero evaluation to be equivalent to a single node.

marattagin also gives some hard data.

you are falsely assuming euler's intent when saying 'absolutely certain' theres a thing called rhetoric.  

txgyc your entire stance is based of taking obscure measurements, info, taking obscure operations on them, and peddling them off as solidly based and concrete logic.  

why else do you think nobody at stockfish has made any of your claims?

Still having a problem understanding big numbers, I see.  Super GMs can't even keep move orders from their own morning prep straight, by the way.  

very interesting, good job Hyvee Huddle App

i clicked to quote on the wrong message i meant the one that u were responding to