Chess will never be solved, here's why

Sort:
playerafar

I'd prefer 'tablebases' with total possible checkmate positions.
That looks like a much more viable project.
Start with two Kings.  None.
Two Kings plus minor piece.  None.

With seven pieces - the number of possible checkmate positions there is a subset of the supposedly 'solved' database (irksome that they skipped castling and en passant possibilities there - very possibly such simply complicates the programming too much - as in slowing it down by too many factors.  But they could or should improve the programming to encompass that)   
Anyway - very posssibly the number of possible checkmate positions with 7 pieces on the board is already precisely known.

playerafar

The math of checkmate positions:
64 places for the white King.  Always a max of 60 for the black King.
With three pieces on board - both the 64 and the 60 shrink tremendously in the case of checkmate.
Suddenly - there's only 28 possible positions for each checkmated King.
And only a maximum of 9 possible positions for the other King in each case.
But maybe - even for just checkmate positions - it could be mathematically shown that the number of possible such positions and the difficulties of isolating-counting-itemizing them are Prohibitive as far as today's programming is concerned.  

playerafar

"Or come with nonsense like insisting on 1 d4 a5?"
a5 enables a 'rooklift' that most other moves would not.
That rooklift Ra6 followed by lateral motion of that rook could be significant in some positions.  

Does Ra3 come in - in the Austrian attack sometimes?  I think so.  
When you're talking about 'solving'  - opinionated 'nonsense' verdicts don't work.  Again - this isn't a courtroom.
Idea:  people who have worked through a lot of tactics puzzles will know about the inadvisability of 'dismissal' when it comes to chess moves.

Mikhail Tal - arguably the greatest tactician ever.
You'd be amazed at some of his masterpiece moves ... not 'studies'. 
Actual GM game rated tournament moves !   Winning moves.
Reactions:   'What ???   How ???'   
Not 'nonsense'.   The Real Deal.  

playerafar

'only one supporting 'my' opinion'
that Svesh was 'right' ...
indeed 'he' might be the only one supporting his own opinion ....

playerafar
Contenchess wrote:

Chess is about mistakes so a computer solving Chess has no bearing on humans. We will still play Chess and we will continue to make mistakes.

Great post !

DiogenesDue
akbolon wrote:

“Have humans walked on Mars? No
Can humans walk on Mars? Yes
Will humans walk on Mars? Maybe, it depends on somebody paying billions of $ to build and launch a spacecraft.”

Let's let Elon Musk manage that...

But I think chess is "solved".

It doesn't matter what you think.  By the agreed upon definitions of solved games, chess is not solved.

DiogenesDue
Optimissed wrote:

Maybe he's just saying that it's a draw.

My post would still apply in that case.

If one's position is that they think that chess is a draw with best play, then just say that.

"I think that chess is a draw with best play."

The words "I" and "think" being the most critical.  It saves 800 pages of back and forth and a man having to flee his own country to live in the Philippines wink.png...

haiaku
tygxc wrote:

I cannot present a peer reviewed publication that states chess can be solved in 5 years, but neither are there any peer reviewed publications that say it takes 5000 years.

Agree.

tygxc wrote:

This is a discussion forum. If we can only post statements published in peer-reviewed publications, then the forum would be empty.

Agree, one can say whatever he wants, but

tygxc wrote:

The facts support the big bang and relativity theory and chess being weakly solvable in 5 years.

Looks like you think your theory is well supported as big bang theory or relativity, even if no peer reviewed paper has been produced to support it.

tygxc wrote:

So far I am the only one who has presented facts and figures supporting my opinion that Sveshnikov was right.
Others just state their opinion.

If you say you have just an opinion, I respect it. I don't know what exactly you mean by "facts". Everything is a fact: the fact that you claim that chess can be solved in 5 years is a fact itself. If by "facts" you mean "evidences" like in a court of law, I can even agree to some extent. If you mean that the facts you provided are all scientific evidences, I disagree and figures per se do not prove anything either: if input is not true, the output will likely be not true too. However, in our discussion in another thread I have, like you, provided figures that in my opinion do not support your claim. Others have done the same here. You think the opposite and I respect your opinion, but people lose interest in arguing with someone endlessly, when no real progress can be made; if not for the OT posts, this thread would likely be already dead.

Vihas_M

The best and the most accurate moves always change according to the opponent. If people will manage to develop a 4000 ELO chess engine, it would be always better than a 3000 one. But only until a stronger engine comes... There's no such maximum level of analyzing to engines although humans have. The further it develops, the better it analyses. 

playerafar


'takes 5000 years' ...  much much more than that !
Unless they have a Very big breakthrough in programming.

There's an aspect of the difficulty not yet mentioned.
The number of piece situations - as opposed to the number of ways in which the pieces in each situation can be placed on 62 squares.
62.  Not 64.
At first glance there might not seem like many piece situations ...
but its actually Huge.

There would be a big 2-dimensional array of numbers - 
with 486 terms in each column and 486 terms in each row.
(that's if  its generous - and doesn't factor in promotions happy.png)
Each term would be the number of ways to place all the pieces for its material situaton on the 62 available squares.  
Then the total number of possible positions would be the sum of those
236,196 terms.    

Derivation of '486' here:
Nine situations with the pawns for one side ...  zero up to eight
times three for the knights (none or one or two)
times three for the bishops
(generous - opposite colored only - if there's two onboard of the side being counted because of disallowing promotions for now)
times three for the rooks
times two for the queen (one or none)   

9x3x3x3x2.

So ...   486 situations for each side ... but that has to be squared
no pun intended
...
before even considering how many possible placements on 62 squares - in each of the 236,196 cases !
Uh oh !

Vihas_M

And if you manage to make an engine can analyze all the 236, 196 positions, that would be the highest elo a possible ( infinity ) until the number of squares or the pieces are increased  

Vihas_M
Vihas_M wrote:

And if you manage to make an engine can analyze all the 236, 196 positions, that would be the highest elo a possible ( infinity ) until the number of squares or the pieces are increased  

* that can

* possible 

tygxc

#1392
"I don't know what exactly you mean by "facts"."

Fact: Weakly solving chess = providing an ideal game with proof that the moves are optimal

Fact: An upper bound for the number of chess diagrams without excess promotion is 3.8521× 10^37.
Fact: Some positions with excess promotions like 4 queens rarely happen in a reasonable game with reasonable moves, thus possibly in an ideal game with optimal moves, but are excluded from the 3.8521× 10^37 count. 
Fact: Inspection of randomly sampled positions included in the 3.8521× 10^37 reveals they cannot be reached by reasonable games with reasonable moves and thus cannot be reached in an ideal game with optimal moves see #861.
Example:

 



Conclusion: 3.8521× 10^37 is an overestimate for the number of positions that can be reached in an ideal game with optimal moves.

Fact: Pawn moves and captures are irreversible in chess. 
Fact: Each pawn move and each capture renders huge numbers of positions not reachable.
Fact: Checkers has been solved using the square root of the number of positions.
Fact: Losing chess has been solved using the square root of the square root of the number of positions.
Conclusion: it is plausible that in weakly solving chess only the square root of positions needs visiting.

Fact: 500 ECO codes A00 to E99 form a partition of chess
Fact: 200 ECO codes B00 to C99 relate to 1 e4
Fact: If can be proven that the 19 ECO codes C67 C65 C54 C53 C52 C51 C50 C49 C47 C45 C34 C33 C29 C28 C26 C24 C22 C21 C20 draw, then it is proven that black can draw against 1 e4
Conclusion: it is only necessary to look at 50 of the 500 ECO codes to weakly solve chess

Fact: cloud engines reach 10^9 nodes / second

General conclusion: sqrt (10^36) / 10^9 * 50/500 = 3.7 years

I would appreciate if you could summarise your own supporting facts from the other thread you mention, or give the link to that post of yours.

Elroch
tygxc wrote:

#1392
"I don't know what exactly you mean by "facts"."

Fact: Weakly solving chess = providing an ideal game with proof that the moves are optimal

This is NOT the definition.  Weakly solving requires the ability to play optimally with either side against _any_ opposing play.

Fact: An upper bound for the number of chess diagrams without excess promotion is 3.8521× 10^37.

False. This is the number of positions WITHOUT PROMOTION. Can't you even read the TITLE of the paper you purport to rely on?
Fact: Some positions with excess promotions like 4 queens rarely happen in a reasonable game with reasonable moves, thus possibly in an ideal game with optimal moves, but are excluded from the 3.8521× 10^37 count. 

Fact: So are positions with a single promotion to queen. The number is useless.
Fact: Inspection of randomly sampled positions included in the 3.8521× 10^37 reveals they cannot be reached by reasonable games with reasonable moves and thus cannot be reached in an ideal game with optimal moves see #861.

Fact: your view of what is reasonable is worthless to a proof.  I would suggest for example that you would not instinctively believe it is reasonable to maneuver pieces for over 500 moves without any capture (and with no pawns on the board) to achieve a win
Fact: If can be proven that the 19 ECO codes C67 C65 C54 C53 C52 C51 C50 C49 C47 C45 C34 C33 C29 C28 C26 C24 C22 C21 C20 draw, then it is proven that black can draw against 1 e4

C20 is 1. e4 e5. Rather than thinking the "reasonable" lines are most of the analysis, you need to deal with all 27 second moves by white, eg 1. e4 e5 2. a4. 
Conclusion: it is only necessary to look at 50 of the 500 ECO codes to weakly solve chess

Irrelevant noise. A00 is virtually as complex as chess itself, as it covers 14 of the 18 legal first moves! It is also likely that none of these moves lose and that they have the same theoretical standing as any respected line (of course this hypothesis is part of what needs to be checked).

 

haiaku

You repeated already, many times, which are those facts. I say I don't know what is the meaning of the word "fact" to you, and if by "facts" you always mean "scientific evidences", I do not agree with you. For example (as @Elroch pointed out, preceding me):

tygxc wrote:

Fact: If can be proven that the 19 ECO codes C67 C65 C54 C53 C52 C51 C50 C49 C47 C45 C34 C33 C29 C28 C26 C24 C22 C21 C20 draw, then it is proven that black can draw against 1 e4

is not a scientific evidence, it's an hypothesis, so you jump to

tygxc wrote:

Conclusion: it is only necessary to look at 50 of the 500 ECO codes to weakly solve chess

The other

tygxc wrote:

Conclusion: it is plausible that in weakly solving chess only the square root of positions needs visiting.

is another hypothesis for the

tygxc wrote:

General conclusion: sqrt (10^36) / 10^9 * 50/500 = 3.7 years

I have already posted, back in this thread, the link to the other thread I mentioned. If you are interested, you can find it by yourself.

It seems to me you pretend to not understand, in order to debate endlessly, maybe to get company, or because you are pleased when people spend their time debating with you. But this is just an hypothesis.

tygxc

#1398
What Sveshnikov said was his opinion, but as he was highly knowledgeable in chess analysis everybody should at least consider what he said instead of outright dismissing it.
It is my firm opinion that Sveshnikov was right indeed.
I presented facts and figures to support my opinion.
So far I have not read any 'perfectly good arguments' which show me to be mistaken.
People seem unable or unwilling to grasp even simple concepts.
I have to repeat myself as long as these do not sink in.
It is even more difficult as the relevant posts get buried under layers of off topic junk and red herrings.
I welcome any valid counterarguments but so far none have been presented.

"the assessment algorithms contained within Stockfish are sufficient to support a weak solution" I have stated several times that the evaluation algorithm is not used for the proof: it is the table base that provides the correct evaluation draw/won/lost when it is hit. The evaluation function could be as simple as a count of material P=1, N=B=3, R=5, Q=9. It only helps to guide the search. The real evaluation is draw/win/loss when hitting the table base, or the calculation ends earlier from 3-fold repetition or stalemate.

The man with the math degree insists on 1 d4 a5 being relevant. If 1 d4 Nf6 is proven a draw, then it is irrelevant if 1 d4 a5 draws or not. If 1 d4 Nf6 cannot draw, then 1 d4 a5 certainly cannot draw either. 1 d4 a5 is nonsense indeed.

The man with the math degree rejects 10^37 positions because promotions can happen. I have several times explained that only excess promotions are excluded and that for each reasonable position with say 4 queens that is not counted there are several positions that are counted but cannot be reached by a reasonable game with reasonable moves and hence cannot be reached by an ideal game with optimal moves. Thus 10^37 is an overestimate. It has been conjectured that only 1 in 10^6 positions is reasonable, can result from a reasonable game with reasonable moves. That would lead to 10^31. Something like 10^36 is on the safe side.

The weakest link in my argument is probably the square root. For checkers is was the square root. For losing chess it was the square root of the square root. It is plausible that for chess it is the square root too, but it might be less or it might be more.

Elroch
tygxc wrote:

The man with the math degree insists on 1 d4 a5 being relevant. If 1 d4 Nf6 is proven a draw, then it is irrelevant if 1 d4 a5 draws or not. If 1 d4 Nf6 cannot draw, then 1 d4 a5 certainly cannot draw either. 1 d4 a5 is nonsense indeed.

This is a guess.

There is no difference between solving chess and any other mathematical theorem - absolute precision is required and guessing carries no weight. This is not debatable.

Thus if 1. d4 is part of your strategy for white, all 18 responses by black are essential to the solution, just like in checkers. 

 

tygxc

#1402
Sorry, no, this is no guess. This is pure logic.
There are only 2 mutually exclusive possibilities: 1 d4 Nf6 either draws or does not draw.
If 1 d4 Nf6 draws, then it does not matter if 1 d4 a5 draws as well or not.
If 1 d4 Nf6 fails to draw, then it is certain that 1 d4 a5 cannot draw either.
"Chess is the art of logic, like music is the art of acoustics" - Karpov
"Chess is a very logical game" - Capablanca

StormCentre3

Elroch lives in an imaginary world whereby - anything is possible.  He leaves out that part - the human mind’s capability to formulate rational thinking-  or Logical Thinking. 

A belief system labeled as Science- irrefutable…. can’t be questioned.

haiaku
StormCentre3 wrote:

A belief system labeled as Science- irrefutable…. can’t be questioned.

Every "system" of thinking is based on beliefs. Even logic is based on postulates. So every system is in fact a theory. Scientific theories are questioned, otherwise we would have never accepted relativity, or quantum mechanics. Mathematics can be questioned too, otherwise we would have never accepted non-Euclidean geometry, for example. We simply think that some postulates are more "evident" than others. Sometimes a postulate becomes more evident a posteriori, when the predictions of the theories based on that postulate are confirmed by experiments.