Chess will never be solved, here's why

Not even this page, never mind "the forums".

I think it does not use the value 9 for a queen. Even after the queen is on the board it rates it 8.x.

This is strange as winning a pawn in 2 or 3 moves is trivial. But it is worth remembering it is adjusting with other positional factors (activity of kings, pawn control of the centre). 

All complete nonsense - it's surprising the thing can play chess at all.

The point is that the valuations are fluid, and that engines' valuations are off by a factor that completely precludes Tygxc's premise when applied at that scale.  The number of pre-filtered eliminated positions by mistaken criteria would be gigantic, and the number of misevaluations of chosen positions would also be gigantic.  You might as well march a million duckling babies over a stormdrain grating and then count the 937,000 you have left and declare it a million ducklings, problem solved.

I'm sure no engine uses integer values for any pieces other than pawns in the starting positions anymore.  Engine improvements are sometimes derived from specific ideas, but often are just tweakings of values that are experimented with and found to work against other engines.  In that sense, engines are simply evolving to play better against other engines and human players, not to play chess perfectly.  Only when they hit the tablebase do they consistently achieve perfect play.

@tygxc I was speaking of chess GAMES, not chess positions... It's actually also amazing how high just the number of positions is, 32 soldiers on 64 squares can create a number that human beings could hardly imagine without some grand scale comparison.

"Missing the point" fallacy.

Quoting out of the context and missing the point.

First of all, it is an epistemological matter.

As above.

Both are exhaustive. See the last section in this post.

Missing the point.

In the second an third section of my previous post I addressed the problems with that statement. As for the reliability, an overgeneralization from a small sample of positions and games has a high probability to be proven false. We can make the following example (maybe someone else here already made it or a similar one, I do not recall well): we have 10²⁰ balls in a basket, few red, the other white; we pick randomly 10000000 balls; the red balls are 1/100000000000 (10⁻¹¹) of the total; the probability p to pick only white balls is approximately p = 0.99999999993. If we do not know a priori that there are red balls in the basket, by inductive reasoning we could well conclude that there is no red ball at all.

Computer-assisted proofs can be affected by glitches, but those proofs can be reproduced, there are methods to verify them, the probability that a glitch occur must be multiplied by the probability that the occurrence is crucial to the proof... In sum, they are much more reliable.

@Optimissed spoke of "full paradigm" for chess, but besides the fact that the word "paradigm" can be used in different ways, if the paradigm is the product of inductive reasoning, it has the same limits of the inductive reasoning.

Overgeneralization and missing the point.

Overgeneralizations. And in that position my Lc0, after 25 million nodes (which is not bad for "her", as you know), still gives White 3.9% chances of winning.

In order to logically use knowledge A to prove something, you have to show that A is already proven; to prove A you can use logic and other proven knowledge B, and so on, down to axioms. Axioms are taken as always true, logic is taken as always true, so if you use them to prove something, that thing is taken as always true. That's what an exhaustive proof is: it is valid in any case. Unfortunately, the knowledge you plan to use is not always true and it is not proven that it would be, with optimal play, because optimal play has not been determined yet. The alpha-beta algorithm is always true: it is proven that it produces the same results of a minimax, so you can use it to cut off lines as soon as the search reaches the endgame. But knowledge as "a bishop is worth 3 pawns" is not always true, it is not exhaustive, so one can use it only for move ordering, when solving chess, not to cut off lines.

Who will prove they understand numbers the best? I will have to keep reading to find out!

Forget about the Observer's Paradox! This is a contest among men to see who has the biggest, most powerful brain. Silly pawns. Your worth is fluid at best.

Overgeneralization.

This is up there with Fermat's Last Theorem, and I am waiting for an Andrew Wiles to appear with proof. Not sure I can wait 350 years though

Incorrect. Sarcasm is the correct definition.

Ironic sarcasm, given that you are a 4-day old troll account afraid to just stand on your reputation, yet talk about contests of "men" (which is sexist, but never mind that).  A category you have excluded yourself from.

#3951

"You do understand that all the point valuations you are using are approximations"
++ I am not talking about point evaluations.
There are only 3 objective evaluations: draw, win, loss.
A pawn is enough to win a game of chess by queening it.
Anything less than a pawn is not enough to win a game of chess as you cannot queen it. Anything more than a pawn is enough to win a game of chess, by trading it for a pawn and then queening that pawn. 

"all other values are fluid" ++ Does not matter, as long as all the other exceed the value of a pawn they can be traded for a pawn that can queen.

"The value of a tempo"
++ The value of a tempo in the initial position is 1/3 of a pawn, not enough to queen, not enough to win. The value of 2 pawns in the initial position is 2/3 of a pawn, still not enough to queen, not enough to win. The value of 3 tempi in the initial position is worth a pawn: the 3 tempo suffice to force the win of a pawn and then queen it.

"the value of every other piece" ++ Does not matter as long as it is > 1 pawn.

"the value of the pawn itself" ++ Does not matter as long as it can queen.

"What is the value of the pawn at a2 in this position?"
++ Does not matter. Can it queen? That is all that counts.

"how long did you want each "relevant" position to be analyzed for again...?"
++ 17 seconds on a 10^9 nodes/s cloud engine, that is 17000 seconds on a desktop.

{Comment deleted, spam: DS}

#3964
"I was speaking of chess GAMES, not chess positions"
++ The number of chess games lies between 10^29241 and 10^34082
https://wismuth.com/chess/longest-game.html 

There are even many trillions of games that lead to 1 e4 e5, e.g.
1. e3 Nf6 2. Bc4 e6 3. Nc3 Nc6 4. Nf3 Bb4 5. Ne2 Qe7 6. Nfd4 Nb8 7. Nb5 Qd8 8. Nbc3 Ng8 9. Ng1 Bf8 10. Nb1 Be7 11. Bf1 Bf8 12. e4 e5

You are planning to have the entire processor array/server farm of your cloud engine evaluate each position one at a time?

What are you planning on using?  Google Compute Engine?  Heroku?  Engine Yard?

You do know that none of these platforms will allow you to run assembly language for a *real* attempt, right?  Nor will they allow you to install an engine directly on each server.  So, your premise is that you will install Stockfish on a virtual machine or something and treat the entire slice of the cloud you are renting out as if it was one giant PC?  

I'm just trying to get a full picture of how you actually think this is going to work...

Reported.  Go sell your wares on an appropriate platform.

What are you trying to say?

That is _NOT_ a general rule. It is highly position dependent. Even the positions with no other material fall into two classes (and the non-winning one is not small). Likewise rook endings, and others. And in general positions, pawn sacrifices are common, even as early as move 2, they can be sound (1. e4 e5 2. f4), or at least plausibly so (the value of this position is uncertain like that of the starting position).

So all your rule needs to become useful is a way to figure out which of the 10^40 or so positions it is true for and which it is false for.

Shoo, run along now.  As if you understood diddly about PaaS platforms...

There we are.

For experiment, one side (let's say is a perfect player who wants the game ends in a draw for this purpose) could build all his/their/her strategy around the fact that two Knights alone can't win.

That alone proves is a draw.

Now imagine all possible draw fortresses there exist evaluated from +3 to +7 that ends in a draw, take R+P vs Q endgame just for an example to not mention other.

The resources for a draw for a perfect player are abundant.

Now imagine how broad just is a road that leads to a draw for a perfect player.

Let that sink in.

The method of elimination eliminates it all.