Chess will never be solved, here's why

@Kptshmot is correct. We cannot be sure this is not so. @tygxc fails to understand the limitations of woefully inadequate empirical information.

Curiously, he has expressed the view that if there were more games like those in the ICCF WC, one or more of them would likely be won. And then he thinks he can draw definite conclusions from the fact that all of them are drawn.

Like my earlier question he needs to explain EXACTLY how many draws in a row it was that changed the result of chess from being uncertain to being definitely a draw. I think the answer is "42", but I just need to see if he agrees. Obviously a mere 41 draws would not provide certainty, but 42 should surely do it.

[Just in case, I am jesting!]

Firstly, a tablebase encapsulates a huge number of mathematical proofs. Each step of its construction uses logic to add to the set of facts you know about chess. For example, suppose you have a position where black is to move where you have previously deduced that black will get mated in 548 moves with optimal play. Retrogade analysis from this position provides a set of positions where white has a mate in 549 moves by playing a specific move. Each of these is a mathematical fact.

Secondly, it is surprising that you have never noticed that if you look at a position in a tablebase it tells you what each move will achieve against optimal defence. This is plenty to tell you the best move to play!

@11367

"he has expressed the view that if there were more games like those in the ICCF WC, one or more of them would likely be won" ++ But by random external factors: clerical error, illness...

"how many draws in a row it was that changed the result of chess from being uncertain to being definitely a draw" ++ Each draw contributed to more certainty, but the absolute certainty comes from the compelling arguments +1 tempo = +1/3 pawn is not enough to win, and each further move dilutes the advantage.

"draw definite conclusions from the fact that all of them are drawn"
++ We certainly can conclude something. In the Second World War the Allies even concluded how many tanks Germany had from the serial numbers of captured tanks.

@11366

"considering all errors equal"
++ An error (?) is a move that worsens the game-theoretical value by 1/2 point.

"Very few available winning lines from the starting position"
++ What could such a hypothetical winning line look like?
1 e4: 3 responses 1...e6, 1...c5, 1...e5 draw.
1 d4: 2 responses 1...d5 and 1...Nf6 draw.
1 Nf3: 2 responses 1...d5 and 1...Nf6 draw.

"engines of similar strength playing each other"
++ ICCF is not engine vs. engine.
It are 17 different human ICCF (grand)masters from 11 different countries who qualified by winning in preliminaries, semifinals, candidates.
Their engines and their settings like contempt are different.
The time per move is different: average 5 days/move, but the humans are free to set more e.g. 10 or less e.g. 2.
The Russians run inferior hardware because of sanctions, but 4 of the 17 finalists are Russian.

"The first error is extremely likely to occur and lead to a draw"
++ So you say chess is a white win, all 17 ICCF (grand)masters make exactly 1 error as white and none of the 17 ICCF (grand)masters can spot the error. Now that is extremely unlikely.

@11371

"The sort of things that could also explain the failure to win a position" ++ No. Clerical error or illness or hasty move lead to immediate defeat: resignation on the next move.

"engines are very imperfect as of today" ++ Yes, but ICCF finalist + engines at 5 days/move have now reached perfection.

"Chess is too big for them to be thorough" ++ Not really. The human looks beyond the horizon. Even the engine quickly hits the 7-men endgame table base as can be seen in TCEC.

"Certainty is a boolean property" ++ When I make it boolean draw/nodraw you come with Bayesian probability 99.999999999999999999999999999999999% sure. When I follow you in your Bayesian probability you come back to boolean.

It is disrespectful to compare an engine assisted correspondence game of 2 years to a coin flip.
Anyway, if I flip a coin 116 times and it lands heads 116 times, I am certain the coin is loaded.
In Bayesian probability the uncertainty is 2^-116 = 10^-35.

"There is a range of uncertainty, but not a range of certainty." ++ This is semantics.

Arrogance and ignorance are an annoying combination. They combine to preserve ignorance. This is a huge advantage of humility, which permits improvement in understanding.

@11373

"Arrogance and ignorance are an annoying combination"
++ 'Everything that irritates you about others is the secret to self-awareness' - Jung

"This is a huge advantage of humility" ++ You could use some of it.

tygxc you still cant provide evidence for any of your claims lmfao

"Anyway, if I flip a coin 116 times and it lands heads 116 times, I am certain the coin is loaded.
In Bayesian probability the uncertainty is 2^-116 = 10^-35

this is false in many ways LOL. first, this makes an assumption of the chances that the coin could be loaded in the first place - for example, after one flip, your 'calculation' would assign a 50% chance of it being loaded, which is ridiculous. your second error is calling 99% certainty.

"When I make it boolean draw/nodraw you come with Bayesian probability 99.999999999999999999999999999999999% sure. When I follow you in your Bayesian probability you come back to boolean."

you do neither lmfao you dont follow either, you are the one who cant switch because you dont have a grasp on the basic logic to do either.

"The Russians run inferior hardware because of sanctions, but 4 of the 17 finalists are Russian."

You do realize that the interview you cite explicitly states that the time period where they had worse equipment they did considerably worse? you are citing a lack of hardware during a time period not covered in the interview that you're citing.

"So you say chess is a white win, all 17 ICCF (grand)masters make exactly 1 error as white and none of the 17 ICCF (grand)masters can spot the error. Now that is extremely unlikely."

treating them all as independent probabilities is hilariously incorrect.

i wonder when it will get through to tygxc that "extremely unlikely ≠ impossible"

""There is a range of uncertainty, but not a range of certainty." ++ This is semantics."

it's literally one of the core errors you make. you treat high likelihood as mathematical fact, and claim a range of certainty. it's not any type of certainty. it's an unknown value.

""draw definite conclusions from the fact that all of them are drawn"
++ We certainly can conclude something. In the Second World War the Allies even concluded how many tanks Germany had from the serial numbers of captured tanks"

ah yes because someone in the past making a statistical inference justifies you claiming statistical inference as fact? LMFAO do you hear yourself right now???? this guy thinks that because some people made an estimate of how many tanks there were, he can just lie, make up statistics, and then call those statistics facts!

it's funny, both elroch and i predicted tygxc would ignore his question, and would you look at what happened! it's really telling how there are some questions, despite being posed at this point dozens of times, that tygxc refuses to answer.

it's really a great sign of intellectual dishonesty.

tbf it's hard to make progress when one party quite literally has no mathematical understanding and is narcissisticly confidently incorrect.

Nobody needs qualifications to comment here.

Certain things are being crassly ignored here about the consecutive draws.
1) similiar engines being played against each other
2) engines with similiar 'horizons' and the limitations of those horizons
3) engines being unable to 'punish' each other's imperfect play relevant to the time given per move
4) engines being unable to win because of issues with how they're programmed to play for a win versus a draw
-----------------------------------------
This all means that a 'concession' is being made that shouldn't be being made.
That so many draws 'increases the probability' that 'chess is a draw with optimal play by both sides'
There's no evidence its optimal play.
None.
Whatever the totally unknown probability is that 'chess is a draw with optimal play by both sides'' the consecutive draws by the engines Doesn't Change It.
So why is that concession being made to T to the contrary?
Is it so he makes more mistakes?
Its not a good trade.
That Concession needs to be interfered with.

Pretty sure such concessions haven't been made, those points have been mentioned before and disregarded afterwards by Tygxc.

@Kotshmot I see posts 'conceding' that the consecutive draws make the overall 'draw' 'more likely. I'm not in a mood to quote them though.
Not looking for a 'courtroom' situation.
If you're holding that the consecutive engine draws don't At All make it 'more likely' 'chess is a draw with so-called Optimal play by both sides' ...
then you and I are in agreement.
There is no 'Optimal play by both sides'. That's a kind of mythical dinosaur that will never exist until chess is thorougly 32-piece tablebase solved.

This evidence is actually very weak. We don't know how good the players are on an absolute scale. (We do know that they are not perfect, but empirically quite good compared to more imperfect players!). There is probably some very difficult way to view the games as evidence and strengthen belief in the value of the game, but not very strongly.

There are games where one side has a forced win but that win is very difficult to find. Chess could be such a game, and if the precision required is similar to that to achieve a win in known tablebase positions, it could be utterly beyond current human and engine combinations.

Take for example KQRkqr endings, a class of equal material endings. The first player to move wins 67.74% of the time, but some of the wins are very long and difficult - max 117 moves to mate.

@Kotshmot
Ironic that Elroch just quoted a 'Concession' post.

There is no 'Optimal play by both sides throughout a game'. That's a kind of mythical dinosaur that will never evolve into real existence until chess is thorougly 32-piece tablebase solved.

I also quoted that post earlier. I'd rather put that post in a different category from concession but fair enough.

Maybe 'concession-assist'.
Like 'assists' in a basketball game.