Chess will never be solved, here's why - Chess Forums - Page 285

MARattigan

Jan 22, 2023

0

#5681

tygxc wrote:

Some more trolling going on here, trolls accusing each other of trolling and trolls putting their ownpersonal definitions above the established definitions.

Prof. van den Herik was a professional game theorist and he defined:

Ultra-weakly solved means that the game-theoretic value of the initial position has been
determined,
weakly solved means that for the initial position a strategy has been determined to achieve the game-theoretic value against any opposition, and
strongly solved is being used for a game for which such a strategy has been determined for all legal positions.
The game-theoretic value of a game is the outcome when all participants play optimally.

According to the trolls he was wrong.

GM Sveshnikov was a professional chess analyst, who taught upcoming masters how to analyse chess with computers. He said:

'Give me five years, good assistants and the latest computers
- I will bring all openings to technical endgames'

According to the trolls he was wrong despite facts and figures confirming he was right.

Why you should worry about Prof. van den Herik's definitions is mystifying because the only thing you do with them is quote them. Your proposals ignore them completely,

I've already pointed out the flaw in the definitions.

Firstly, the Allis definition that Prof. van den Herik quotes for "weakly solved" (what you promise to do) requires no limit on the time required to apply the strategy. If you want to solve, say, pre or post 2017 basic rules chess (with, say, the resignation and agreed draw rules excised to render them soluble) then according to that definition you're too late. Syzygy has already solved them (for any number of men - he's even given you a quick access tablebase for most relevant positions less than 8 men).

(That's not a problem from a game theory point of view, but obviously not appropriate for a solution in terms of OP's question.)

Secondly the definition doesn't say if the strategy is for one player or both. If, for example, the initial position happens to be a win for White and also forced selfmate for Black then a strategy for Black to mate himself would count as a solution according to that definition. But not according to the man in the street. If Black doesn't follow the strategy (which would obviously be his most sensible course of action) then it's just back to a normal (unsolved) game of chess.

Thirdly, if the initial position is a draw then a strategy for one player that achieves a win against some opposition and a draw against the rest would not count as a solution for that player; the strategy would need to provide a draw whatever. That would also be in conflict with the man in the street's idea of what constitutes a solution.

As for your facts and figures, you yourself are perfectly aware they confirm nothing of the kind, so your reference to trolls accusing people of trolling is at least partly apt.

MARattigan

Jan 22, 2023

0

#5682

haiaku wrote:

Careful. According to Allis', van den Herik's and other definitions, a weak solution provides the game-theoretic value of the initial position and a strategy to achieve at least that value. ...

My reading of Allis', van den Herik's definition is that the strategy provides that value, not at least that value.

The intention was probably at least that value, but the wording is not.

And the definition also doesn't call for the game-theoretic value of the initial position, only the strategy to achieve the value (which needn't be known).

DiogenesDue

Jan 22, 2023

0

#5683

Optimissed wrote:

Facts and figures that no-one here has agreed with and that many here have accurately criticised. It's a toss-up as to whether you or MAR are troll of the century.

I've been using computers to help analyse chess for 40 years. The principles are extremely easy to work out for oneself. Mainly that we must always work out fail-safe tests for computer analysis, by intervening to force continuations.

That's an exaggeration. While there was chess game software in the early to mid-80s, the "analysis" they could provide was below master level. Significantly so.

The period when computer analysis could be helpful for blunder checking and poking and prodding openings and endgames and sometimes finding obscure middle game tactics started later, and ran until the 90s. Once Deep Blue level analysis became mainstream, though, that all changed. When Kramnik lost to an engine in 2006 in the last serious even-odds challenge between human and engine there has been, that was the tombstone on that era.

You've disparaged Stockfish many times, and seem to think we are still in the era where engines could not find Greek sacrifices (for example) or anything past a horizon of 20-30 moves. The positions that engines have trouble with now are more and more fringe positions that would not occur in a human game. That being said, assuming engines are perfect and can evaluate perfect play is garbage. It's just that humans are pretty easy to beat, and engines have left them behind and are still going forward at a pace that is still measured in mere weeks. It's inherently obvious that if a team can make an adjustment to their engine mid-TCEC tournament and achieve a significant difference in results that engines are *nowhere near* perfect play.

When Stockfish release 80-something cannot be improved upon for years at a time, *then* there will be some kind of argument that engines are reaching a "draw roadblock" similar to super GMs that choose known drawish lines. The fact that humans can reach state such a state among themselves but still be massacred by engines proves that similarly, when engines start to draw more and more often, that trend can adequately be explained by engines reaching their own playing peak under current technology...it doesn't mean that chess itself is a draw. If aliens landed tomorrow and started wiping the floor with chess engines, that would *still* not prove chess is a draw.

DiogenesDue

Jan 22, 2023

0

#5684

MARattigan wrote:

Why you should worry about Prof. van den Herik's definitions is mystifying because the only thing you do with them is quote them. Your proposals ignore them completely, [snip]

This is how Tygxc props up his premise...the same way that flat earthers prop up their ideas by referring to real science when it support a narrow point they wish to make.

All the calls to authority he makes are bogus. On Tromp's thread somewhere, you can watch this process directly in action, as Tygxc attempts to tack his ideas onto Tromp's analysis and push Tromp to make statements, etc. Tromp politely declines and ultimately tells Tygxc his notions are scientifically invalid. If Tygxc had had access to Sveshnikov or van den Herik, he would have been dismissed similarly..."thanks for being a fan of my work, but no, your premise doesn't hold water". This false attachment to reputation is what allows Tygxc to pretend he somehow is right when everybody is else is wrong and is crucial to his ability to continue to say the exact same unsupported things 100 times over in the face of literally everybody more in the know than beginners opposing him. Without the name-dropping, he would just look like a crackpot immediately.

DiogenesDue

Jan 22, 2023

0

#5685

Optimissed wrote:

At first there was hardly any read-out. You had to hop back and forwards, letting it come up with moves, making the move, seeing what it came up with. It was very difficult. Before I had a P.C. running Dos, I had a Sinclair QL, which wasn't bad and about the size of a laptop which needed plugging in to a TV. But I knew people with other computers from the 70s. It was difficult to use computers for analysis but quite possible.

The IBM PC came out in 1981. I played using every chess software I could find on pretty much every platform around long before and after this. Apples, Ataris, Amigas, PCs. I had all of them at my house, and many other more obscure personal computers besides (not to mention playing chess on VAXs and PDP11/70s which obviously were not at my house). My rating at the time would have been about 1300 I would say. There was no software I was not capable of beating on its highest setting until Chessmaster came out in 1986 (and that was a serious leap, as level 3 (out of about 10) was the level I could comfortably win about 50-50. So chess analysis via software was worthless prior to that, and even with the initial release of Chessmaster, it would be most useful only for basic blunder-checking. It certainly would have been entire incapable of, say, redeeming the Berlin ala Kramnik's analysis for the 2000 WCC.

This video may be of interest here:

It lists the very best engines in 1985 at 1900 rating (and that is UCSF rating, I believe, not FIDE, so call it 1800 maybe). It was 1996 when the first (affordable/accessible) chess engine broke 2600 and could be considered on par with top GMs of the time. It was 2006 when a chess engine hit 3000 rating and effectively put human players in the rearview mirror.

llama36

Jan 22, 2023

0

#5686

MARattigan wrote:

the definition also doesn't call for the game-theoretic value of the initial position, only the strategy to achieve the value (which needn't be known).

Well... it needn't be known in the case that you can win, because if you can force a win then sure, you've solved it.

But let's say we come up with a strategy that always draws in chess... we couldn't call that a weak solution without knowing a better result isn't possible.

llama36

Jan 22, 2023

0

#5687

MARattigan wrote:

llama36 wrote:

MARattigan wrote:

the definition also doesn't call for the game-theoretic value of the initial position, only the strategy to achieve the value (which needn't be known).

Well... it needn't be known if it's a win, because if you can force a win then sure, you've solved it.

But let's say we come up with a strategy that always draws in chess... we couldn't call that a weak solution without knowing a better result isn't possible.

I'd agree with the first.

If you come up with two strategies, one for each player that always draw then you've also solved it.

And if you come up with two strategies, one for each player that always at least draw then you've solved it again.

Sure, if you have two independent unbeatable strategies, one for each player, then you know the game theoretic result... maybe in some cases that would be a way to approach the problem of solving a game.

(And I guess for this we're assuming the game has only 3 outcomes, ranked best to worst as win, draw, loss)

MARattigan

Jan 22, 2023

0

#5688

llama36 wrote:

MARattigan wrote:

llama36 wrote:

MARattigan wrote:

the definition also doesn't call for the game-theoretic value of the initial position, only the strategy to achieve the value (which needn't be known).

Well... it needn't be known if it's a win, because if you can force a win then sure, you've solved it.

But let's say we come up with a strategy that always draws in chess... we couldn't call that a weak solution without knowing a better result isn't possible.

I'd agree with the first.

If you come up with two strategies, one for each player that always draw then you've also solved it.

And if you come up with two strategies, one for each player that always at least draw then you've solved it again.

Sure, if you have two independent unbeatable strategies, one for each player, then you know the game theoretic result... maybe in some cases that would be a way to approach the problem of solving a game.

(And I guess for this we're assuming the game has only 3 outcomes, ranked best to worst as win, draw, loss)

As for the last, yes, FIDE need to fix their laws so that is the case and the results for the opponent occur in reverse order. (Without a definite ranking the phrase "at least a draw", for example, makes no sense.)

If you have a strategy that wins for just one player then as you said you have a solution. No strategy is required for the other.

My suggested definitions correspond with what you say for the drawn case.

haiaku

Jan 22, 2023

0

#5689

MARattigan wrote:

haiaku wrote:

Careful. According to Allis', van den Herik's and other definitions, a weak solution provides the game-theoretic value of the initial position and a strategy to achieve at least that value. ...

My reading of Allis', van den Herik's definition is that the strategy provides that value, not at least that value.

The intention was probably at least that value, but the wording is not.

And the definition also doesn't call for the game-theoretic value of the initial position, only the strategy to achieve the value (which needn't be known).

Allis' definition of weakly solved is:

For the initial position, a strategy has been determined to obtain at least the game–theoretic value of the game, for both players, under reasonable resources.

The game–theoretic value is the highest outcome that can be achieved by force and a strategy suited for a weak solution must be proven to achieve the game–theoretic value of the initial position, from the initial position itself. I concede that Allis' definition is somewhat redundant and using "at least" may be superfluous (it is not possible to achieve by force a value higher than the game-theoretic one, and values not achieved by force do not guarantee that the strategy can achieve the — possibly lower — game–theoretic value).

I too used "at least", though, to emphasize that it is not necessary that an optimal strategy for a weak solution always achieves the game–theoretic value of the current position in a game. This value is higher, if the opponent does not play optimally, than the game–theoretic one at the start; so an optimal strategy, that has only to obtain the game–theoretic value of the initial position, may not be the best, when the opponent makes mistakes that change that value.

MEGACHE3SE

Jan 22, 2023

0

#5690

Optimissed wrote:

MEGACHE3SE wrote:

Optimissed wrote:

no thats actually the official stuff, I have known that since before this discussion

Do you mean it's what can be found on Wiki when Googling? I did have the misfortune to be coerced into reading the so-called official stuff and I thought it was so bad and generally out of focus that it needs to be challenged. When I first saw it, I genuinely thought it was a joke, devised by some philosophy professor or other, just to see who would believe it. My subject's philosophy and some of the errors would fail a second-year undergrad essay.

It's mish-mash of improperly related ideas, probably put out by people calling themselves "games theorists" and who probably don't know what games theory really is. Actually, the general definition of games theory, to be found on Wiki, isn't too bad. It's about an application of games STRATEGY to real life situations, to turn those situations into a game simulation model which can be scored; scores being applied to outcomes and jiggled about until something approaching real life outcomes is achieved. Then that model can be applied to different strategies of dealing with those r. l. situations. Obviously, a game itself can become the object of a games theoretical approach, with alternative strategies being scored. This will be the source of the general confusion in the so-called official definitions, especially regarding the specialised meaning of "strategy".

This is the main problem with the practice of referring to the product of a solution of chess as a strategy, because to anyone who understands the ideas involved, assigning scores to positions is the strategy programmed into chess engine algorithms. Whereas the product of a solution of chess is wrongly claimed to be a strategy, it really consists of finding accurate moves, in response to moves made by the opponent. A workable strategy isn't going to consist of finding bad moves or anything else than good moves, which are defined as not changing the game state! However, at the moment, none of the algorithms is accurate enough to base a dependable solution on it.

Some of the types of people here buy straight into this pseudo-intellectualised "games theory approach", of course without understanding the basics. Undoubtedly, where they lead, others follow, because such people spend much of their time building a façade of intellectualism, with which to con others.

no its because I literally studied game theory when I was younger for math competitions and this is the official terms.

tygxc

Jan 22, 2023

0

#5691

@7708

"Allis' definition"
++ Allis was the first to define in his thesis on Connect Four.
Van den Herik improved the wording in his paper on solving games.

MARattigan

Jan 23, 2023

0

#5692

haiaku wrote:

MARattigan wrote:

haiaku wrote:

Careful. According to Allis', van den Herik's and other definitions, a weak solution provides the game-theoretic value of the initial position and a strategy to achieve at least that value. ...

My reading of Allis', van den Herik's definition is that the strategy provides that value, not at least that value.

The intention was probably at least that value, but the wording is not.

And the definition also doesn't call for the game-theoretic value of the initial position, only the strategy to achieve the value (which needn't be known).

Allis' definition of weakly solved is:

For the initial position, a strategy has been determined to obtain at least the game–theoretic value of the game, for both players, under reasonable resources.

You are correct: https://cris.maastrichtuniversity.nl/en/publications/searching-for-solutions-in-games-and-artificial-intelligence

The version appearing in the van den Herik paper to which @tygxc refers, and attributed to Allis is not a faithful reproduction.

With Allis' definitions, none of the objections I raised to @tygxc's stated definition in #7692 apply.

The game–theoretic value is the highest outcome that can be achieved by force and a strategy suited for a weak solution must be proven to achieve the game–theoretic value of the initial position, from the initial position itself. I concede that Allis' definition is somewhat redundant and using "at least" may be superfluous (it is not possible to achieve by force a value higher than the game-theoretic one, and values not achieved by force do not guarantee that the strategy can achieve the — possibly lower — game–theoretic value).

I too used "at least", though, to emphasize that it is not necessary that an optimal strategy for a weak solution always achieves the game–theoretic value of the current position in a game. This value is higher, if the opponent does not play optimally, than the game–theoretic one at the start; so an optimal strategy, that has only to obtain the game–theoretic value of the initial position, may not be the best, when the opponent makes mistakes that change that value.

I agree with the above. My objection to dropping the phrase "at least" as in the version quoted by @tygxc was not that it allows solutions that achieve less than highest possible yield against some opposition; rather that it rules out solutions that do sometimes achieve higher yields than the game-theoretic value of the initial position.

MARattigan

Jan 23, 2023

0

#5693

tygxc wrote:

@7708

"Allis' definition"
++ Allis was the first to define in his thesis on Connect Four.
Van den Herik improved the wording in his paper on solving games.

Just what wording did van den Herik improve on from Allis' thesis on Connect Four? I don't believe Allis gave any definiition of "solved" in his thesis.

You could, I suppose, take the view that something is an improvement on nothing, but it does lead to the objections I raised in #7692.

You haven't yet commented on those objections. Do you accept them as valid?

(Not that it has any importance to your project to solve chess of course, because you don't intend to take any notice of any definition of the term, but just out of interest.)

tygxc

Jan 23, 2023

0

#5694

@7712

Victor Allis was a Master student of Professor van den Herik.

MARattigan

Jan 23, 2023

0

#5695

Yes , we know that.

What about the question?

tygxc

Jan 23, 2023

0

#5696

@7712
No, I do not consider 'objections' of @7692 as valid.

The definitions by Prof. van den Herik have no flaw at all.
They are carefully worded and improve on Allis' first version on minor points.

"no limit on the time"
++ That is irrelevant. Solved is solved, whether it takes years, or months, or centuries.

"a solution in terms of OP's question" ++ The original poster did not specify whether his question was about ultra-weakly, weakly, or strongly solved.

"the definition doesn't say if the strategy is for one player or both"
++ It does: the game-theoretic value is when all participants play optimally.

"If, for example, the initial position happens to be a win for White and also forced selfmate for Black then a strategy for Black to mate himself would count as a solution"
++ No, the definition calls for all opposition. The selfmate does not oppose to the win.

"if the initial position is a draw then a strategy for one player that achieves a win against some opposition and a draw against the rest would not count as a solution for that player"
++ If you can win, then you can draw as well. This issue was treated in the solution of Losing Chess, which is a win for white.
During the solution they considered draws and wins for black as failures for white.

"the man in the street's idea" ++ The man in the street should read van den Herik's paper to understand the difference between ultra-weakly, weakly, and strongly solving.

MARattigan

Jan 23, 2023

0

#5697

tygxc wrote:

@7712
No, I do not consider 'objections' of @7692 as valid.

I meant to refer to #7691, but you appear to be addressing that post anyway.

The definitions by Prof. van den Herik have no flaw at all.
They are carefully worded and improve on Allis' first version on minor points.

You give above no reasoned argument against my objections, just your opinion that Allis' definitions in his peer reviewed paper (presumably the one I posted in #7710 rather than anything in the one you quoted in #7709) are inferior to those in van den Herik's peer reviewed paper, without any justification for that opinion.

You do give arguments against some of the points in my post #7691 below which I'll address individually.

Prof. van den Herik attributes the definitions he gives directly to Allis and the Allis paper to which I referred. He makes no mention of any alteration from the original, so I think we can assume the version appearing in the van den Herik paper is simply a faulty transcription.

My objections in #7691 relate only to the version in van den Herik paper. Allis' definitions are essentially identical to the ones I posted myself.

"no limit on the time"
++ That is irrelevant. Solved is solved, whether it takes years, or months, or centuries.

A solution according to either of the above definitions involves determining a strategy.

You have to distinguish between the time it takes to produce the strategy and the time it takes to apply the strategy.

If there is no limit on time on either then the following strategy for a player is a solution (of FIDE chess variants suitably amended to be soluble).

1. If it is not your turn do nothing, otherwise

2. Determine the game states after each possible legal move.

3. Follow Syzygy's algorithm until each of the game states determined in 2 has been assigned a win or frustrated win for you or your opponent or until no further entries can appear with the material in any of the game states determined in 2. Note that this step does not require any external input from a tablebase.

4. If a game state determined in 2 has been assigned a win for you in 3 play a move that leads to one with the lowest DTZ. Otherwise if a game state determined in 2 has been assigned a frustrated win for you in 3 play a move that leads to one with the lowest DTZ. Otherwise if a game state determined in 2 has not been assigned a win or frustrated win for your opponent in 3 play a move that leads to one. Otherwise if a game state determined in 2 has not been assigned a win for your opponent in 3 play a move that leads to one. Otherwise play anything.

So if you believe that a time limit on producing or applying the solution is irrelevant, you may as well stop posting your offers to solve chess for us. Syzygy has already beaten you to it.

"a solution in terms of OP's question" ++ The original poster did not specify whether his question was about ultra-weakly, weakly, or strongly solved.

This phrase that you've snipped out of the middle of my post was used in the context of time limitation as discussed above. Nothing to do with the type of solution.

Since it's patently obvious that there is a strategy for producing any type of solution given enough time and resources, it can reasonable be assumed that OP was referring to a solution that could be applied in reasonable time with reasonable resources.

"the definition doesn't say if the strategy is for one player or both"
++ It does: the game-theoretic value is when all participants play optimally.

A definition of game-theoretic value may well refer to both players, though it wouldn't usually refer to a strategy.

Why are you talking about that? It should be clear that the definition I was talking about was your definition of "weak solution".

"If, for example, the initial position happens to be a win for White and also forced selfmate for Black then a strategy for Black to mate himself would count as a solution"
++ No, the definition calls for all opposition. The selfmate does not oppose to the win.

Neither does anything else against perfect play.

It's obvious to anybody not trolling that Allis. van den Herik and myself mean, by "against any opposition", simply, "against any legal actions by the opponent". Nobody but yourself is ever going to take as meaning "against what @tygxc thinks are good moves".

"if the initial position is a draw then a strategy for one player that achieves a win against some opposition and a draw against the rest would not count as a solution for that player"
++ If you can win, then you can draw as well.

OK. Either prove that Black can't win in the position below or post your move to draw.

Black to play, ply count 0

This issue was treated in the solution of Losing Chess, which is a win for white.
During the solution they considered draws and wins for black as failures for white.

"the man in the street's idea" ++ The man in the street should read van den Herik's paper to understand the difference between ultra-weakly, weakly, and strongly solving.

He would get a better idea if he read Allis' paper. You've apparently read both but understood neither.

tygxc

Jan 23, 2023

0

#5698

@7717

Allis wrote in his 1994 PhD Thesis promoted by Prof. van den Herik:

ultra-weakly solved For the initial position(s), the game-theoretic value has been determined.
weakly solved For the initial position(s), a strategy has been determined to obtain at least the game-theoretic value of the game, for both players, under reasonable resources.
strongly solved For all legal positions, a strategy has been determined to obtain the game-theoretic value of the position, for both players, under reasonable resources.

His promotor Prof van den Herik wrote in his 2002 paper:

Here ultra-weakly solved means that the game-theoretic value of the initial position has been determined,
weakly solved means that for the initial position a strategy has been determined to achieve the game-theoretic value against any opposition, and
strongly solved is being used for a game for which such a strategy has been determined for all legal positions

The latter paper is later, and is written by the professor, not the student.
It improves on the 'at least' and 'for both players' with the wording 'against any opposition'.
It does away with the vague 'reasonable resources'.

Elroch

Jan 23, 2023

0

#5699

At least you can now see that "DETERMINED" does not include "Ignore this position because we can make something up on the fly for any continuations arising from it". "Determined" entails "specified".

If you can't I don't know why you haven't claimed you have already solved chess because you can make up a way to draw on the fly (since chess is "obviously" a draw).

MARattigan

Jan 23, 2023

0

#5700

tygxc wrote:

@7717

Allis wrote in his 1994 PhD Thesis promoted by Prof. van den Herik:

ultra-weakly solved For the initial position(s), the game-theoretic value has been determined.
weakly solved For the initial position(s), a strategy has been determined to obtain at least the game-theoretic value of the game, for both players, under reasonable resources.
strongly solved For all legal positions, a strategy has been determined to obtain the game-theoretic value of the position, for both players, under reasonable resources.

His promotor Prof van den Herik wrote in his 2002 paper:

Here ultra-weakly solved means that the game-theoretic value of the initial position has been determined,
weakly solved means that for the initial position a strategy has been determined to achieve the game-theoretic value against any opposition, and
strongly solved is being used for a game for which such a strategy has been determined for all legal positions

The latter paper is later, and is written by the professor, not the student.

So you think your big brother is bigger than my big brother. Well whoopidoo.

Your big brother's paper states that the definitions are from my big brother's paper.

They're not. They're an inaccurate transcription of the definitions in my big brother's paper. The inaccuracies give rise to the objections I raised in #7691.

It improves on the 'at least' and 'for both players' with the wording 'against any opposition'.

That is the opposite of improvement. That is the reason for my second and third objections in #7691.
It does away with the vague 'reasonable resources'.

That again is the opposite of improvement. That is the reason for my first objection in #7691.

It would also mean you can stop peddling your offer to solve chess because Syzygy has already beaten you to it.

You haven't produced any pertinent points in rebuttal of #7691. Is claiming your big brother is bigger than mine the best you can do?