Will computers ever solve chess?

The lunatics are running the asylum now?

Let Alpha Zero play against itself.  Then do Monte Carlo simulations on the emerging results.   Simple.  Except someone has to pay for the computing time.

Of course you might need to do this exercise for (God Knows How Long) but at least you would conceptually be "making progress" towards a more fuller understanding of any "forced outcome" to the game.

Sounds better than 7,000+ posted dominated by Symbolic Logic 101 exercises.

P.S., the "Saint Lunatics" are a great Rap / Hip Hop Band from Saint Louis, Missouri, with lead singer "Nelly."

Neural network engines use Monte Carlo Tree Search as the actual search engine.

MCTS + Bit Boards + Magics are the main components of the actual game engine.

So what is the point of the after the fact simulations?

You answered your own question.  The simulations are apparently not needed.

Thanks for the clarification, (assuming you're correct).

The neural network is just used to sort the list of possible moves for MCTS to further optimise the search engine.

MCTS starts with just random moves and its results are used to supervise the machine learning.

Wanna know about tail end gradient descent man?

I do it my way, not their way.

Computers cannot solve chess

Yes they can.

Nope.

[Yawn].  Alpha Zero won @20 games out of 100 (with about 80 draws) to win against the current engine champ.  A preponderance of draws dovetails with a thread (from a couple years back) by the former US Correspondent Champ.  He argued in his tread that chess was likely a draw with best play by both sides.  Why pinheads find his conjecture so outlandish is what's behind the silliness running throughout this interminable thread.  7000 posts later, this thread still reads like a lot of one-legged-men in an ass-kicking contest.

Onward Christian Soldiers.  Knock yourselves out debating the meaning of "solved."  

Pigs CAN fly.  Just rent an airplane.

Thanks for the unwanted attention @Manatibrain, and have a nice day.

P.S. -- I like the way you edited out you loose-cannon temper, and wide-ranging penchant for invective (in your email below).  Neat trick.

What are the chess.com definitions of "Best" move and a "Brilliant" move? Would the chess.com engine be able to find and report a "Brilliant" move? Are there other ways to make moves better than the chess.com "Best" move?

lol: no, it's not impossible. The horizon effect is that analysis only extends a certain distance. It may be overturned by a big surprise at some later point in the game (or multiple related surprises in different lines). This can happen at any depth, although sometimes it's just too far for an engine to ever reach. I have looked at positions (quite simple endings) that engines calculate to depth 100 and still get wrong. If this depth was increased further and further, eventually it would get it right in principle (it may be impractical in terms of computing time and space).

Engines playing themselves would be expected to have more draws because they disagree with other engines more about the value of a position, which means there is more chance of being proven wrong in the latter case.

LMAO.....

The best resource on the combinatorics of chess endgames. Just needs a little work to extend it to the full 32 piece tablebase.

https://kirill-kryukov.com/chess/nulp/results.html

Selected facts:

• there are exactly 38,176,306,877,748,245 legal positions in an 8 piece tablebase (this has not yet been generated in full)
• there are 29,045,304 possible material balances that can be reached in a game of chess (just specifying the number of pieces of each of the 5 types each side has. Just 5 because each definitely has one king!)

I beat a 39 y.o. Fidelity last night while drinking bourbon and reading a book.

Price tag is a cool 400 million man.

Do you still think that this is all about a game?

Space & Time