Chess will never be solved, here's why
Sort:
TITAN3338998
Aug 31, 2022
0
#4361
Just tell me one thing.If the best chess computers in the world say to repeat moves are the moves then why are you suggesting a move which is clearly worse and not that best?For me in a perfect game of chess only the best moves are played and not the second best.
Aug 31, 2022
0
#4362
There are different concepts which are referred to as "best", but the key one is that a second-best move leads to a worse result to the best move, assuming table-base optimal play thereafter.
Aug 31, 2022
0
#4363
I fully agree with you.This is why I am wondering why someone here is thinking that you can play a different move here?
Aug 31, 2022
0
#4364
Good morning,
I have a question:
Let’s say that there is a better shot for the whites at the first shot (if there is one?) and that the blacks and whites then play each time the best shot possible, it implies that it could be the perfect part and therefore that if they repeat another part and they replay each time the best moves, that this second part should be similar to the first. For this it is necessary that the first blow of the whites is identical to the first part. But how to determine that the first blow of the whites is 100% the best blow that exists?
Isn’t it the best shot that exists according to the response of blacks to this first shot?
Sorry, I may have deviated from your discussion, but I’m interested in understanding the logic of the best shots from the very beginning of the game, since the rest of the time depends on those first shots and mainly on the first shot of the whites.
Have a good evening
tygxc wrote:
@4428
"You need to consider all the others"
++ No, the onus of proof is on the one who claims a win with equal material.
The onus of proof in on the one who claims a draw with unequal material.
Incorrect. Until you have reached some kind of theoretical threshold (and a dead GM's one line supposition is not it), you cannot make any assumptions. Even if your hypothesis is believed by many at a particular point in time ("the earth is clearly flat" or "the earth is clearly not flat"), if you are going to follow the scientific method, then you are not at liberty to simply declare one position over the other no matter how likely you think it is. Thank goodness you aren't in any position of making any actual scientific determinations, because your bias makes you the worst kind of scientist (well, barring one who makes pathogens for fun).
So, once and for all time...the onus is on you. This argument evolved from the original discussion about whether chess is actually a forced draw, and you are arguing as if the two sides of this argument are:
- Chess is draw
- Chess is not a draw
...but that is not what is being argued. The two consensus positions are actually:
- Chess is "known" to be a forced draw (Ponz, Tygxc, Optimissed)
- Chess is probably a forced draw, but no claims of certainty can be made
The latter position reflects current reality. So, the onus must therefore always fall on the former to make a proof of their "knowledge". The fact that arguments like "anyone knows that XYZ, or you are just a bad chess player" are being bandied about is a good indicator of the hollowness of that position.
You also seem to misunderstand how engines work. They reach their evals by making assumptions of value based on their evolved valuations of chess (human-derived valuations for traditional engines, computer derived for newer engines, or a hybrid of the two...all are imperfect play) and pruning massive amounts of lines thereby. This is why it is still possible for ICCF players to "suggest" different lines and suddenly come upon one that the engine values more highly after deep evaluation of the suggested move, where the engine had summarily discarded it before.
Because engine analysis is flawed, and can't be used as proof of perfect play.
Elroch wrote:
There are different concepts which are referred to as "best", but the key one is that a second-best move leads to a worse result to the best move, assuming table-base optimal play thereafter.
I don't see that, since there are three game results possible. A second best move may lead to the SAME result as the best one.
@4454
"the key one is that a second-best move leads to a worse result to the best move, assuming table-base optimal play thereafter."
++ No. There may be and usually are several good moves, that lead to the same result.
Elroch wrote:
Optimissed wrote:
The word is "refuted". Falsification is different. Falsification is all about the theoretical possibility that something can be refuted. If something cannot be falsified, that means that something isn't based on scientific evidence. Therefore there's no reason to believe that it's true in any case, so it doesn't need to be disproven.
The belief that 2. Ba6 doesn't necessarily lose isn't based on any evidence and therefore it can be discounted. That's because assumptions work both ways and these people calling those who are sure it loses "incorrect" are therefore incorrect themselves. This discussion shouldn't be happening, because the assumption that we cannot tell that Ba6 loses is only based on indirect evidence, perhaps that "something else" may not lose although we think it does.
Basically, Elroch is criticising us for using inductive evidence but his evidence is equally inductive. So he loses.
Actually, I am not criticising anyone for using inductive evidence. The problem is incompetent use of it, directly analogous to the reasoning "I have drawn 10 white balls from this urn, therefore the next ball I draw will be white". Also, when I use inductive reasoning, I don't disguise it, nor do I ever claim certainty. Certainty arises from deductive reasoning (such as the mathematics of Bayesian probability. Theorems are applied and conclusions follow.
Anyone who still has fully functioning cognition should be able to follow the following consideration of what can be achieved with inductive reasoning:
Suppose we want to conclude that the position 1. d4 Nf6 2. Bh6 is winning for white. We start from the time when we have just learnt the rules of chess by have no knowledge of the game. (Anyone who thinks that at this point we have certain knowledge that the position is a loss for white is clearly delusional, so I hope we can agree that at this point the value of the position is uncertain, i.e. has a probability strictly between 0 and 1 of being a win for white). Take this time to be when chess had just been invented, if you like.
From that point we (the human species, assisted by computers) acquire evidence. This comes in units which we will break down right to the position level - empirical data about what happens in chess games.
At every step in the long path to wisdom by experience we start with a state of belief about whether the opening position is a win or not. This is quantified as a Bayesian probability - a state of belief. After an element of evidence, we have another state of belief about the proposition, quantified as another probability.
The issue is that those who believe we can be genuinely certain about the result of 1. d4 Nf6 2. Bh6 are obliged to say that at some time, the observation of a single new position - eg the end of a game with a result - caused an uncertain belief about the proposition to change to a certain one.
This is absurd except in one single case - when the last position completed a deductive proof that the result of the position is a win (just like the final step in the proof of the result of checkers, or the step that adds a new position to a tablebase, with a certain result (by all of the positions reachable from it in one move also being in the tablebase).
The idea that one elementary piece of inductive evidence could otherwise justify a change from uncertainty to certainty is misguided and simply wrong.
Of course, I understand that the likes of @tygx will simply ignore this truth and go on as before - the ability to improve understanding can be too limited.
[Note, this description is also a close approximation to the way an AI learns chess. At each point in time it has a model with millions of parameters and it revises those parameters based on any mismatch between its probabilistic model and experience.
It would be possible to repeat the development of AlphaZero with all games starting from the position after 1. d4 Nf6 2. Bh6. This would generate a chess player able to play both sides of this position better than anyone else on the planet. This player would still never be certain about the result of the position of interest (barring issues with machine precision).]
Not claiming certainty is not really meaningful. It's self-protection in a lot of cases, where we actually can be certain of something, due to inductive evidence.
I explained to you why I thought that Alpha Zero is actually responding to its own programming. When it gives a probability for 2. Ba6 as beng winning which is not actually unity, that doesn't mean that the position may not be winning for black. It's for the following reason.
I tried to explain why and in what way AlphaZero is not programmed perfectly. It's because it cannot be, in the same way that a human cannot fully analyse chess to a definite result. Now, the position after 2. Ba6 is not probably won for black. That's because, theoretically, it is either a definite forced win for black, a definite forced win for white, or a definite draw with best play by both sides.
All AlphaXero is doing, when it gives that probability, is giving an estimation of its own error limits.
Sep 1, 2022
0
#4369
Elroch wrote:
... "I have drawn 10 white balls from this urn, therefore the next ball I draw will be white"....
But only mathematicians draw balls from urns. Other people get their fingers covered in ashes or burnt by hot tea.
Sep 1, 2022
0
#4370
Nicoquelicots a écrit :
Good morning,
I have a question:
Let’s say that there is a better shot for the whites at the first shot (if there is one?) and that the blacks and whites then play each time the best shot possible, it implies that it could be the perfect part and therefore that if they repeat another part and they replay each time the best moves, that this second part should be similar to the first. For this it is necessary that the first blow of the whites is identical to the first part. But how to determine that the first blow of the whites is 100% the best blow that exists?
Isn’t it the best shot that exists according to the response of blacks to this first shot?
Sorry, I may have deviated from your discussion, but I’m interested in understanding the logic of the best shots from the very beginning of the game, since the rest of the time depends on those first shots and mainly on the first shot of the whites.
Have a good evening
Well, I see that my question is of no interest to anyone.
Hello the intelos egos who will still blablabla for 200 pages to know who is right! !
Good luck to all!
Sep 1, 2022
0
#4371
Optimissed wrote:
Not claiming certainty is not really meaningful.
It is, regardless of whether you understand that it is. And acknowledging uncertainty is crucial in the real world, where uncertainty is the norm. The question of the value of chess is an interesting case because (unlike much of the real world) it could hypothetically become certain - it is practical issues that make it uncertain (and likely to stay so for many years).
I explained to you why I thought that Alpha Zero is actually responding to its own programming.
AlphaZero is not programmed beyond the level of an imbecilic beginner playing entirely random legal moves. All its expertise is learnt.
When it gives a probability for 2. Ba6 as beng winning which is not actually unity, that doesn't mean that the position may not be winning for black.
Correct.
It's for the following reason.
The reason is that it cannot prove a specific result. Nothing more.
I tried to explain why and in what way AlphaZero is not programmed perfectly.
Again with that "programming". It is an AI, not a program.
The reason AlphaZero is imperfect is its finite (though large) size. The other limiting factor is the amount of learning, but this is not the limiting factor.
It's because it cannot be, in the same way that a human cannot fully analyse chess to a definite result. Now, the position after 2. Ba6 is not probably won for black. That's because, theoretically, it is either a definite forced win for black, a definite forced win for white, or a definite draw with best play by both sides.
It is wildly unlikely that the position is won for black. I will pedantically and correctly avoid making the mistake of claiming it is a solid fact, but you can feel safe to do so (the probability of being proven wrong is extremely low).
All AlphaXero is doing, when it gives that probability, is giving an estimation of its own error limits.
This is like claiming that Carlsen's statements about the value of different moves are "giving an estimation of [his] own error limits".
They ain't.
Sep 1, 2022
0
#4372
This forum is only about if Chess will ever be solved or not!Not about something else if you have a question to something else go in that forum and not in this one and stop writing nonsense which no one here cares about!!!
Sep 1, 2022
0
#4373
@Nicoquelicots
I didn't understand the connection between the line, "Isn’t it the best shot that exists..." and anything in the para., "Let’s say that there is a better shot for the whites at the first shot...", which is why you got no response from me. Probably the same for everybody else.
Perhaps you need to elaborate.
Nicoquelicots wrote:
Nicoquelicots a écrit :
Good morning,
I have a question:
Let’s say that there is a better shot for the whites at the first shot (if there is one?) and that the blacks and whites then play each time the best shot possible, it implies that it could be the perfect part and therefore that if they repeat another part and they replay each time the best moves, that this second part should be similar to the first. For this it is necessary that the first blow of the whites is identical to the first part. But how to determine that the first blow of the whites is 100% the best blow that exists?
Isn’t it the best shot that exists according to the response of blacks to this first shot?
Sorry, I may have deviated from your discussion, but I’m interested in understanding the logic of the best shots from the very beginning of the game, since the rest of the time depends on those first shots and mainly on the first shot of the whites.
Have a good evening
Well, I see that my question is of no interest to anyone.
Hello the intelos egos who will still blablabla for 200 pages to know who is right! !
Good luck to all!
Your English is probably difficult for some of them to understand. Generally the best moves at the beginning of a game are those which facilitate movement of your pieces and which control centre squares. The game is all about controlling squares and communication between pieces. As the game progresses, you can start to work out which squares you want to control more and which pieces need to be positioned better.
Hope that helps a little.
Elroch wrote:
Optimissed wrote:
Not claiming certainty is not really meaningful.
It is, regardless of whether you understand that it is. And acknowledging uncertainty is crucial in the real world, where uncertainty is the norm. The question of the value of chess is an interesting case because (unlike much of the real world) it could hypothetically become certain - it is practical issues that make it uncertain (and likely to stay so for many years).
I explained to you why I thought that Alpha Zero is actually responding to its own programming.
AlphaZero is not programmed beyond the level of an imbecilic beginner playing entirely random legal moves. All its expertise is learnt.
When it gives a probability for 2. Ba6 as beng winning which is not actually unity, that doesn't mean that the position may not be winning for black.
Correct.
It's for the following reason.
The reason is that it cannot prove a specific result. Nothing more.
I tried to explain why and in what way AlphaZero is not programmed perfectly.
Again with that "programming". It is an AI, not a program.
The reason AlphaZero is imperfect is its finite (though large) size. The other limiting factor is the amount of learning, but this is not the limiting factor.
It's because it cannot be, in the same way that a human cannot fully analyse chess to a definite result. Now, the position after 2. Ba6 is not probably won for black. That's because, theoretically, it is either a definite forced win for black, a definite forced win for white, or a definite draw with best play by both sides.
It is wildly unlikely that the position is won for black. I will pedantically and correctly avoid making the mistake of claiming it is a solid fact, but you can feel safe to do so (the probability of being proven wrong is extremely low).
All AlphaXero is doing, when it gives that probability, is giving an estimation of its own error limits.
This is like claiming that Carlsen's statements about the value of different moves are "giving an estimation of [his] own error limits".
They ain't.
Sometimes, I'm afraid I get the feeling that you don't know what you're talking about.
Is what you say understood by yourself due to inductive evidence? If so, how certain are you that you're correct?
The thing you wrote about Carlsen was largely correct and you used it as an analogy of bad thinking. You would benefit by quite a lot if you could learn to understand how others think. Particularly those who are cleverer than you are. 
I'll try to explain it. You're a statistician and therefore accustomed to dealing with sets of events in the "macro". Considering a unique event, you will tend to see its outcome "a" or outcome "b" probabalistically, since as a unique event but still an event which is viewed in the context of the set, superficially outcomes a or b follow probability patterns. Within the context of the set, viewing all the events as a whole, each unique event is simply part of this or that subset as part of a statistical pattern. That tells us nothing about the event itself.
An outcome on the chess board may be considered to have a definite result with best play: but it is one that may be unknown. All probabilities that are assigned to it are therefore the result of guesswork based on inductive reasoning. That is effectively equivalent to basing them on error limits.
Anyway, that's my thinking. You can't win an argument just by telling me that you're right.
You made a logical error in your paragraph beginning with It is wildly unlikely that the position is won for black.
You may wish to edit it. Personally, I don't believe there is any difference between the probabilities that the position is won for white and a draw. They are both probabilities of zero. I can't begin to guess why you assign them different probabilities. A bit like the other week, when Man City was winning by a massive number of goals with about a couple of minutes to play. The "probability forecaster" was still giving a probability of about 3 in ten thousand that it would be a draw and yet one in ten thousand that it would be a win for the other team. If you think about it, there's a reason why that is logically incorrect, as well as being wildly unlikely. Therefore the numbers were just there for show. A bit like my IQ really, which is so far out of the normal scale that the figure is meaningless, because the sample size is very small. When you have an IQ of only 31, like I do, it's completely meaningless.
Sep 1, 2022
0
#4377
Optimissed a écrit :
Nicoquelicots wrote:
Nicoquelicots a écrit :
Good morning,
I have a question:
Let’s say that there is a better shot for the whites at the first shot (if there is one?) and that the blacks and whites then play each time the best shot possible, it implies that it could be the perfect part and therefore that if they repeat another part and they replay each time the best moves, that this second part should be similar to the first. For this it is necessary that the first blow of the whites is identical to the first part. But how to determine that the first blow of the whites is 100% the best blow that exists?
Isn’t it the best shot that exists according to the response of blacks to this first shot?
Sorry, I may have deviated from your discussion, but I’m interested in understanding the logic of the best shots from the very beginning of the game, since the rest of the time depends on those first shots and mainly on the first shot of the whites.
Have a good evening
Well, I see that my question is of no interest to anyone.
Hello the intelos egos who will still blablabla for 200 pages to know who is right! !
Good luck to all!
Your English is probably difficult for some of them to understand. Generally the best moves at the beginning of a game are those which facilitate movement of your pieces and which control centre squares. The game is all about controlling squares and communication between pieces. As the game progresses, you can start to work out which squares you want to control more and which pieces need to be positioned better.
Hope that helps a little.
Thank you for your answer. I am not a native English speaker and I am doing my best.
I know that shots at the very beginning of a game are the most important to have the best control of the parts. But in all the theories I read in the messages of all of you, you never mention the first moves when it is precisely these that determine the whole sequence of a game. So, to go to the end of the playdoyer of all of you to know the perfect shots, I ask the question again:
"What is the perfect first blow, that is, that of the whites?"
And I think there is no perfect answer because the first shot of the whites can only be considered perfect after the first shot of the blacks, depending on what was played. It’s a bit like asking the question: who’s first, the chicken or the egg?
No?
Nicoquelicots wrote:
Optimissed a écrit :
Nicoquelicots wrote:
Nicoquelicots a écrit :
Good morning,
I have a question:
Let’s say that there is a better shot for the whites at the first shot (if there is one?) and that the blacks and whites then play each time the best shot possible, it implies that it could be the perfect part and therefore that if they repeat another part and they replay each time the best moves, that this second part should be similar to the first. For this it is necessary that the first blow of the whites is identical to the first part. But how to determine that the first blow of the whites is 100% the best blow that exists?
Isn’t it the best shot that exists according to the response of blacks to this first shot?
Sorry, I may have deviated from your discussion, but I’m interested in understanding the logic of the best shots from the very beginning of the game, since the rest of the time depends on those first shots and mainly on the first shot of the whites.
Have a good evening
Well, I see that my question is of no interest to anyone.
Hello the intelos egos who will still blablabla for 200 pages to know who is right! !
Good luck to all!
Your English is probably difficult for some of them to understand. Generally the best moves at the beginning of a game are those which facilitate movement of your pieces and which control centre squares. The game is all about controlling squares and communication between pieces. As the game progresses, you can start to work out which squares you want to control more and which pieces need to be positioned better.
Hope that helps a little.
Thank you for your answer. I am not a native English speaker and I am doing my best.
I know that shots at the very beginning of a game are the most important to have the best control of the parts. But in all the theories I read in the messages of all of you, you never mention the first moves when it is precisely these that determine the whole sequence of a game. So, to go to the end of the playdoyer of all of you to know the perfect shots, I ask the question again:
"What is the perfect first blow, that is, that of the whites?"
And I think there is no perfect answer because the first shot of the whites can only be considered perfect after the first shot of the blacks, depending on what was played. It’s a bit like asking the question: who’s first, the chicken or the egg?
No?
Thankyou very much for your answer.
I know that there's something in what you are saying. I personally completely agree with you. There is no "best move". However, there are some moves which may be less helpful than others, such as, perhaps, 1. Na6.
Sep 1, 2022
0
#4379
Because on the contrary, if there is really a perfect blow of the whites from the first shot, it would mean that if we could find all the perfect shots, also from the first blow of the blacks. We could then invariably play the same game, with 100% success for whites or losers, since their first shot is perfect. I hope that we will never arrive at this hypothesis because the game of chess would no longer have any interest.
The more we evolve, the more there is of no one, voluntary or not. And frankly, losers are even more frustrating than victories or defeats!