Solving chess? With no BS. (moderated)

This thread is for discussion of the possibilities, or lack thereof, of solving chess in a provable manner. 

Rules:

1. Being 99% certain isn't good enough to make any claim.  Discussion is fine.  Opinions are fine.  Don't make claims that you cannot prove, just based on your opinion.

References:

Shannon's paper on solving chess circa 1950 (how many of the people that are arguing about this topic and using Shannon's number in discussion have actually read this?)

Summary:

This topic is for those who believe in the scientific method and those who aspire to solve chess, or prove that it cannot be solved.  Chess players without any programming background or knowledge of solved games are welcome, but much like a Formula One driver who tries to correct their pit crew mechanics, opinions will be taken with a grain of salt.  Playing a game, in and of itself, does not qualify someone to make claims about whether it can actually be solved. 

Are you Hans Berliner?  Welcome.  Post away.  Are you Ben Finegold?  Do your homework on the subject before you start tossing stuff around (no offense intended, Mr. Finegold ...).

This thread will have far less posts than other threads discussing this topic, because I will be actively removing posters who don't follow the rules, post opinions as facts, make false appeals to authority, etc.  

Definitions:

1. Chess

Chess is defined by the basic rules.  Tournament rules like the 50 move rules, et al are not actually part of the game of chess, they are additional rules for competition (you can discuss any rules you like here, just keep this distinction in mind, because ultimately we're talking about solving the basic game).

2. A game of chess

A game of chess is a logical construct, and is defined as any legal sequence of moves that achieves an outcome of checkmate, stalemate, or draw by insufficient material (pawnlocked positions and perpetual checks are also forced draws and thus fine).  Agreed upon draws are allowed by convention and tournament rules in the interest of time, but are technically incomplete games of chess (so, do not use agreed upon draws as evidence when discussing whether chess can be solved).

3. Chess player

A chess player is any sentient being or any artificial engine or software that is capable of playing a complete game of chess when presented with another player's moves in a format it can understand (so, a correspondence player is the human being, and the engines they use are all separate players...do not refer to human + engines as a single player, they are a team of players)

4. Solved games

5. Tablebase

Premises:

- Chess is still unsolved (by the definition of solved games given above)

- Many of the strongest human players believe that chess is likely a forced draw, but can't be sure and haven't proven it, nor do they claim to have proven it

- Engines surpassed the best human play somewhere between 15 to 20 years ago now, and the top engines are currently rated around 500 to 700 points higher than the best human players

- Engines evaluations are imperfect and engines are still improving with each new release, and alone, cannot be reliable evidence of "perfect play"

- The definition of "perfect play" or "best play" is imprecise and subjective, and is utterly dependent on the judgment/valuations of human beings and engines that cannot play perfect chess. For the purposes of this thread, the only moves that can be called "perfect play" are moves/sequences of moves that are backed up by a tablebase

- There are approximately 10^46 unique chess positions. Estimates by various studies vary, anywhere from 10^40 to 10^50 depending on the rigor applied. The study most referred to as being the best update of Shannon's original numbers is:

https://content.iospress.com/articles/icga-journal/icg19-3-05

Update: John Tromp's study, which has been (somewhat loosely thus far) peer reviewed, brings the number of unque legal positions down to 10^44.6.

- Checkers was solved in 2007, by calculating 10^14 positions, which is 100,000,000,000,000,000,000,000,000,000,000 times less than the number of chess positions that need to be calculated

- Tablebases are slowly "solving" chess , going backwards from checkmate, but practically, have only reached 7 piece tablebases (around 140 terabytes). Work on an 8 piece tablebase is underway (will be over 5 petabytes). To fully solve chess this way, a 32 piece tablebase would be needed, which would require an amount of storage that requires more than all the matter in the known universe currently.

Opinion:

Even if you could record each node on a seperate subatomic particle, there aren't enough particles in the entire observable universe (something under 10^80) to keep track of the nodes in the game-tree (around 10^120 if you only consider the first 40 moves, with no promotions allowed).

Yep.  Even with quantum computing multiplying the number of trackable states for each atom at the sub-atomic level (better than electron charges, direction of spin, etc.), it's impractical to talk about solving chess as things sit currently.  You could scoop up the entire asteroid belt and make a quantum computer out of it and it would probably fall short.

People argue that you can "prune the tree" and get there, but we'd have to "prune" about 99.999999999999999999999999999999% of all legal positions to make a serious dent.  I don't think anyone could seriously say that only 1 out of every billion billion billion moves is worthy of being evaluated as candidates for "perfect play".

We can still talk about what it *would* take, though.

chess will never be solved (Opinion)

You should have labeled that as an opinion. But yeah.

okay, Iw ill edit that comment @blueemu

Lol.  I don't mind opinions being stated without disclaimers (although it's refreshing to see who can distinguish the difference ), as long as some backup is coming if the opinion is challenged.  I just want to avoid the circular arguments that inevitable ensue when somebody keeps saying something they can't actually support...

How can you completely solve the game of chess when the number of chess games is 10^120 which is more than the number of atoms in the observable universe?

That is the topic, yes.  How would you do it?

First efforts have been to reduce to the number of unique positions, not games (10^46), then to posit "pruning" from there, but it falls woefully short.

Are there "discoveries" about chess yet to be made (by engines in all likelihood) that would allow for massive swathes of the game tree to be bypassed?  That seems like one of the only ways to progress...if Moore's Law were to hold up long term (it is not holding up right now, and speed of light issues with silicon wafers are going to keep slowing it down unless we make some new breakthrough), it would still be a long, long, long time before technology will just solve chess by brute force.  Even the current tablebases are not completely brute force, they take shortcuts.

The key is taking shortcuts that can be proven to be 100% safe.  One overly sketchy shortcut and you lose the validity of everything and have to calculate again.

When checkers was solved, the space was 10^20, which was "pruned" to 10^14 to solve the game...if we can only knock off 6 orders of magnitude, the 10^46 number is still way out of reasonable range.

I say that chess is solvable with future tech. It will be one of the things that future people won't be able to play anymore because of people being able to access something like that

(opinion)

In game theory, chess is a zero-sum two-person dynamic game of perfect information. Zero-sum roughly means players have opposing interests in the game. Perfect information means that at any point in the game players know the past moves. Mathematician von Neumann showed that every zero-sum two-person game has a solution. This, however, does not mean that always the player who starts will win the game. Some games have second-mover advantage.
For example, take the Nim 21 game (Nim - Wikipedia

).
It is a dynamic two-person game of perfect information, just like chess. Each player take turns adding a number. The starting number is 0. Each player increases the number by 1, 2, or 3. The player who says 21 or more loses.
Think about the game and confirm that the winning strategy for player 2 is to always say a multiple of 4. No matter what player 1 says (1,2, or 3), player 2 can increase it to 4, and so on. Player 1 will always lose this game.

It can be argued that the 1st move is a psychological advantage only. 
Who’s the fellow who argued it is actually Black with the advantage- since he is the 1st to act on the information  revealed by White?

Great example for those that are convinced the opening move is always an advantage, which should be obvious, but it's not.  I like to use the rock-paper-scissors example, too.  Rock, paper, scissors is a balanced game by design, but if you force one player to select rock, paper, or scissors first, the other player always wins and the game becomes obviously silly.

Is three fold repetition also an acceptable way for a game to end? Coupled with the absence of the 50 move rule, both players could simply keep moving their piece(s) continually without making any progress. 

I argue before any attempt at proving a forced win by White - 1st it is necessary to prove initiative = advantage. Should be a simpler task ?

Otherwise - could not Black simply mimic White forcing a draw ? Or perhaps even win as White exhausts possible moves resulting in zug ?

We know many positions whereby the side to move losses by force.

When I was younger and playing through some closed games by Lasker and Rubenstein, I often wondered if black with best play could force either a pawnlocked closed game (draw) or force white to make a capture that would compromise their game slightly.

I no longer think this is at all likely, but it's not hard to imagine a few rules tweaks that would create this outcome for the game.

Having the initiative does NOT equal having the advantage.

Or perhaps the opposite can be proven. For every example where Black mimics White an advantage is eventually found.