Chess will never be solved, here's why

Sort:
DiogenesDue
MaetsNori wrote:

True. It should, of course, be a draw, due to how symmetrical the starting position is.

Even in 960, the two sides mirror each other.

This is why I've thought it would be far more interesting for a chess variant where each color has their starting position randomized (just as in Fischer Random) ... but each color's arrangement is independent of the other.

So we might have a starting position like this:

Then the players begin.

To make it fair, games are played in pairs, like in the TCEC - where the same setup is used for the next game, but the players switch sides ...

This would certainly reduce the number of draws, I believe ... and would make memorization and theory virtually irrelevant.

I'm sure others have thought of this (or a similar idea) as well ...

I have no issue with your premise. You should fix your example position, though...read up on king and rook placement for Chess960.

MaetsNori
DiogenesDue wrote:

I have no issue with your premise. You should fix your example position, though...read up on king and rook placement for Chess960.

Ah, thanks for the guidance. I had to look it up.

Elroch
MaetsNori wrote:
llama_l wrote:
tygxc wrote:

I am certain chess is a draw with optimal play, and I have justification.

I (and nearly everyone who plays chess) assumes chess is a draw with best play, and they can give reasonable arguments for that.

True. It should, of course, be a draw, due to how symmetrical the starting position is.

Not at all a good reason. And it's not symmetrical! White has the move. There are many games which are as symmetrical (in the loose sense) but which are won by one side. For example, Nim with any set of sizes to start. In this game, the first player wins in most starting positions and the second player wins in a subset of positions with a special property.

Even in 960, the two sides mirror each other.

This is why I've thought it would be far more interesting for a chess variant where each color has their starting position randomized (just as in Fischer Random) ... but each color's arrangement is independent of the other.

Have you not heard that this is one of the more popular alternative variants?

This would certainly reduce the number of draws, I believe ... and would make memorization and theory virtually irrelevant.

Chess960 is plenty adequate for that. You only see the same starting position in 1 in 960 games, so learning a bit of theory about each is probably already scarcely possible!

I'm sure others have thought of this (or a similar idea) as well ...

Good guess!

tygxc

@12647

"Guessing is not good enough." ++ OK, then please take a games data base and tell how many underpromotions, promotions to pieces not previously captured, and underpromotions to pieces not previously captured you can find.

"I referred to generating drawing strategies for a MODIFIED game where underpromotions to bishop and rook are forbidden." ++ That is too restrictive: what I talk about is a modified game where underpromotions to pieces not previously captured are forbidden.

"My statement is true"
++ No 4*10^37 refers to no promotions to pieces not previously captured.

"a weak solution is: it involves generating a complete proof tree."
++ Not necessarily. Counterexample: Allis' weak solution of Connect Four.

"In a proof tree not a single legal move for the opposition is left unanalysed."
++ That is your opinion/interpretation. Mine is that 1 e4 e5 2 Ba6? and 1 a4 can be pruned.

"the game theoretic value (W/D/L) of a position is the maximum of the game theoretic values of the positions reachable by a legal move from that position."
++ The game-theoretic value of the positions reachable by 1 e4 e5 2 Ba6? is lower than those reachable by 1 e4 e5 2 Nf3. The game-theoretic value of positions reachable by 1 a4 cannot be higher than those reachable by 1 e4.

"Current chess engines beat those of only a short time in the past in some games."
++ In short time controls, not in 5 days/move.

"The ICCF is a competition between Stockfish 16 and itself running on high end hardware."
++ No. you have no clue about correspondence chess.
The humans make the difference, not their engines.

tygxc

@12648

"Previous ICCF championships may have had greater disparities between competitors"
++ They had the same cycle of qualifiers.

"ICCF players may have learned to trust their hardware and software more"
++ That is not how correspondence works. The humans trust their engines on tactics, but take care of strategy beyond the horizon of the engines.
'How many ideas can you interactively throw at the computer in one hour is the key question'
Interview with 2 times ICCF World Champion Langeveld

"competitors all pitting the same hardware/software against each other with minimized human interference" ++ No, then they would not win qualifiers.

"ICCF players may have learned to embrace less volatile openings/defenses and have thus narrowed their repertoires to the safest options"
++ No they play risky stuff, like the Catalan, the Najdorf, the Grünfeld, Nimzovich Indian.

playerafar

tygxc would appear to be disagreeing with others what 'weakly solved' is.
As I've said - that terminology is the gremlin at the center of his invalid proposals.

MaetsNori
tygxc wrote:

"ICCF players may have learned to trust their hardware and software more"
++ That is not how correspondence works. The humans trust their engines on tactics, but take care of strategy beyond the horizon of the engines.
'How many ideas can you interactively throw at the computer in one hour is the key question'
Interview with 2 times ICCF World Champion Langeveld

This was an interesting interview, I appreciate the link.

I especially found this part noteworthy: "Ideas need to be created behind the board and then have them refuted by the computer until you find an idea that sticks and makes the difference."

So I'm getting a mental image of an ICCF player asking their engine(s), "Does this work? No? How about this? No? ... Maybe this? No? ... Hmm, what about this? Maybe? Okay, let's analyze this further ..."

If so, that sounds quite a lot, to me, like "trusting their hardware/software". Engine-assisted trial and error, at least ... Perhaps ICCF players aren't following their hardware/software blindly (and credit to them, for that!) - but they certainly sound reliant on it.

This is quite a different approach from a top OTB player (Carlsen, for example), who may glance at an engine line for a brief amount of time, tell his seconds, "I get the idea", then trust in himself to navigate the game from there, should it arise on the board ...

An ICCF player, apparently, does the near-opposite: comes up with some ideas, then trusts in the engine to navigate the game, from there ...

Thee_Ghostess_Lola

10^44, which is the same "hard number" everyone is using.

then prove its right or wrong. insteada just talking & not doing.

Thee_Ghostess_Lola

ohh and btw ?...already know u couldnt show us what u have...cuz u have n/t to show lol ! u probably couldnt even write hello world in basic burst !

tygxc

@12657

"what weakly solved is"
++ Again:
'weakly solved means that for the initial position a strategy has been determined to achieve the game-theoretic value against any opposition' Games solved: Now and in the future

There is a different interpretation about what 'any opposition' means. Elroch thinks it is all legal moves. I say all legal moves that oppose, i.e. strive against achieving the game-theoretic value.

There is a different interpretation about what 'a strategy' means. Elroch thinks only about a brute force method. As per the above paper I also think about knowledge based methods, like Allis' weak solution of Connect Four

tygxc

@12661

"Carlsen plays OTB, where practical elements matter."
++ Even in correspondence practical elements matter. Sometimes they get into time trouble. They have 50 days per 10 moves, but sometimes they use all of it and then must reply the same day for some moves. Sometimes ICCF games get lost by clerical errors: they play the move they planned ahead for the next move. There are also other human elements like illness. They may also experience computer crashes or power blackouts.

MaetsNori
llama_l wrote:

Eh, that's a bit too weird for my taste. IIRC in a recent top level 960 event some of the top players had a bad / lost position before move 5... so I don't think we need to mix it up even more.

The Casablanca thing (?) I think they called it, (and not that's not a typo) had them starting from historical middlegame positions, that was a neat idea.

Yes, perhaps 960 is already difficult enough.

Also I agree, the Casablanca tournament was quite interesting. Though I found myself wondering why they had 3 Supers and one 2600 ... Seemed a little out of place.

DiogenesDue
Thee_Ghostess_Lola wrote:

ohh and btw ?...already know u couldnt show us what u have...cuz u have n/t to show lol ! u probably couldnt even write hello world in basic burst !

[and]

10^44, which is the same "hard number" everyone is using.

then prove its right or wrong. insteada just talking & not doing.

I can write a "hello world" program in assembly language, C++/C#/C, SQL (Oracle or SQL Server), Ruby on Rails, Python, Java, Javascript, PHP(LAMP), Perl, PAL, Unix shell script, batch files...and Basic. I could make it client/server using AJAX or Django.

I could dig out old textbooks and write "hello world" programs in COBOL and Pascal, too, I guess, if I had any motivation to do so.

I can also *build* worlds using the Doom or Quake engines, Neverwinter Nights, etc. and have in-game NPCs say "hello world" if it came to that.

In any case, you haven't been paying much attention...I have my own thread where this question was pretty much laid to rest, so why would I rehash everything over and over again on a thread that should have died long ago?

chessplayerfootballplayer

OHH'

QSO67

"Once chess solver, always cheater." wink.png

MEGACHE3SE
tygxc wrote:

@12645

"it not a useful way to bring 10^43 down to 10^17"
++ Bringing down 10^44 to 10^38 is by eliminating vast numbers of positions like the 3 random samples shown in https://github.com/tromp/ChessPositionRanking. None of these can result from optimal play by both sides, as there are underpromotions to rooks and or bishops from both sides and underpromoting to a rook or bishop only makes sense to avoid a draw by stalemate, and it cannot be optimal play for both sides to avoid the draw.

Bringing down 10^38 to 10^34 is by eliminating positions like

 

10^34 is conservative: Tromp conjectured a reduction of 10^6 i.e. to 10^32.

The reduction from 10^34 to 10^17 is the square root assuming perfect alpha-beta pruning. Schaeffer arrived only at exponent 0.67 for Checkers, but chess programs have evolved more than Chinook, and Chess is easier to prune than Checkers.
The real exponent lies somewhere between 0.5 and 0.67.

"You need to be able to discard entire classes of positions"
++ Yes: after 1 e4 discard all classes of positions with a white pawn on e2.
After 1 e4 c5 discard all classes of positions with a black pawn on c7.
After 1 e4 c5 2 Nf3 d6 discard all classes of positions with a black pawn on d7.
After 1 e4 c5 2 Nf3 d6 3 d4 discard all classes of positions with a white pawn on d2.
After 1 e4 c5 2 Nf3 d6 3 d4 cxd4 discard all classes of positions with 32 men, or 8 white pawns.
After 1 e4 c5 2 Nf3 d6 3 d4 cxd4 4 Nxd4 discard all classes of positions with 31 men, or 8 black pawns.
After 1 e4 c5 2 Nf3 d6 3 d4 cxd4 4 Nxd4 Nf6 5 Nc3 a6 discard all classes of positions with a black pawn on a7.

"any position with 3 bishops for one player can be discarded"
++ The only way to reach 3 bishops is by underpromoting a pawn. The only reason to underpromote to a bishop is to avoid a draw by stalemate. It happens, but it is very rare.
By the time a pawn reaches the back rank normally one bishop has been traded.
It happens that a pawn promotes with 2 bishops still on the board, but it is very rare.
The combination of both unlikely events: underpromotion to a bishop to avoid stalemate,
and a pawn reaching the back rank with both bishops still on the board never occurs.

You misrepresent your samples to justify your fantasy. the position you cite as in the 10^38 number has already been filtered out by the 10^38. number.

"Schaeffer arrived only at exponent 0.67 for Checkers, but chess programs have evolved more than Chinook, and Chess is easier to prune than Checkers. 
The real exponent lies somewhere between 0.5 and 0.67."

this claim is completely unfounded lmfao. In fact, shaeffer literally contradicts you here. maybe you should have read the paper you cite.

MEGACHE3SE

a reminder to all observers that ive brought tygxc's "logic" to dozens of math majors and math professors and all of them found the same errors. in fact, almost all of them chided me for wasting my time with someone as illogical as tygxc.

MEGACHE3SE
tygxc wrote:

@12657

"what weakly solved is"
++ Again:
'weakly solved means that for the initial position a strategy has been determined to achieve the game-theoretic value against any opposition' Games solved: Now and in the future

There is a different interpretation about what 'any opposition' means. Elroch thinks it is all legal moves. I say all legal moves that oppose, i.e. strive against achieving the game-theoretic value.

there is no "interpretation". Elroch is objectively correct while you are objectively incorrect. anybody who isnt laboring under some delusional fantasy can tell that. if it doesnt address all possibilities then it isnt a solution, because those possibilities are by definition unresolved until rigorously dealt with.

There is a different interpretation about what 'a strategy' means. Elroch thinks only about a brute force method. As per the above paper I also think about knowledge based methods, like Allis' weak solution of Connect Four

There is no "different interpretation", you just have a faulty method. You dont use the "knowledge based method", as that method was rigorously proven, while yours isn't. the strategy must be as mathematically rigorous as the brute force by definition. your method is neither brute force nor knowledge based.

why do you continue to cite papers and definitions that you have no understanding of?

MEGACHE3SE

"10^34 is conservative: Tromp conjectured a reduction of 10^6 i.e. to 10^32."

he never did that lmfao.

if you actually READ THE SOURCES YOU CITED you would see that tromp's starting number is 10^44.

the 10^38 comes from an entirely different paper entirely.

Thee_Ghostess_Lola

well if u know sooo much abt programming then just hard solve the first (10) ply of a chess game. lets see what ur made of. pick any language u want. lol !!