Chess will never be solved, here's why

Sort:
playerafar

"a weak solve means only the games derived from perfect play are known"
No perfect game of chess has ever been established to have ever happened.
Ever.
And regarding Elroch's post and 'magic step' 
is claiming 'know more than all of you here' that 'magic step'?
Somebody here (not I and not Elroch) has emphatically so claimed.
We know who it is. Does it explain all his posts?
Very very possibly if not probably. More like probably.
------------------------------
History of the forum:

1) There was a gigantic claim by the 'knows more' person that the basic top speeds of computers in a chess context 'don't matter'.
That one didn't sell very well. Most people would know better. Do know better.
2) Same 'knows more than all' person tried to claim 'taking the square root' of the number of possible legal positions is 'OK'.
Anybody who knows anything about exponents knows its not.
That one didn't 'sell big' either. happy
3) But then some engines kept drawing each other.
That filled our 'knows more than everyone' friend with Much Hope.
Is now his Main Push. Along with the ++ stuff.
With the tactical references back to the failed #1 and #2 when needed so he can keep circling back to Main Push.
4) But various much better informed posters and posters much more free of illogic (big overlap there) then bring out the real information and logic while exposing and refuting the postings of the 'knows more' person and the discussion becomes informative instead of disinformational.
5) A moderator coming in helping the quality of discussion.

Elroch

Here's an example of Bayesian probability.

The proposition is:

47127418247124127428012420880848921392173297412972983482691264127966916284263432862841268461298642198461298461298461249812649126428682687 is prime

I don't know if this is true or false. The number was generated as a sequence of (quite) random digits. I checked the last one wasn't 5 or even, but nothing else.

@tygxc would say there are only two values of this proposition, so you have to either take the view it is true or it is false. This is wrong.

The appropriate Bayesian probability - a quantification of belief - would be the one that makes you neutral about the betting odds. To explain that, if the probability of an outcome is p, the fair odds on that outcome is (1-p) to p. i.e. if you stake p and the proposition is true, you get 1-p profit. If it is not you lose p. To put it another way, if you stake 1 that the number is prime you get (1-p)/p profit for being right and lose 1 for being wrong.

Your Bayesian probability is the one where you are indifferent whether you bet that the proposition is true (with odds (1-p)/p to 1) or false (with odds (p/(1-p) to 1).

Anyhow, back to that possibly prime number.

One approach would be to remember than the density of primes at N is about 1/log(N). So if you worked out the natural log of that number (easy to do approximately) and inverted it, that would be a probability you could use.

But you could improve on that. The last number not being a 5 or even increases the chance it is a prime (all the primes are in the other 4/10 numbers., so you increase p by a factor of 10/4=2.5 ). Then you could add all the digits up to see if the number is divisible by 3. If it is, bingo, you can make p=0! If not, p has increased again, but a factor of 3/2 this time.

An interesting point is that, as a betting game, the person who does a better job of estimating the probability is in a good position. If someone else offers different odds, he believes he knows which side is profitable. This is true even though both he and the other person might have worse estimates of probability than someone with better knowledge.

The best knowledge of all would be to have worked out that the number is prime or composite (This might be too hard for the players, but they would be told later).

playerafar

binary - as opposed to 'unknown' and subject to 'probability' and what probability is.
One of the things I like about these discussions is how the history of math begins to come out.
I was great at math in school. And geography.
But not good at history. Didn't like it either.
Much much later I had much more appreciation of history.
-----------------------------
And it became clearer that if one goes at history through the history of math and science - then its going to make much more sense than prioritising through history of politics/religion.
With geography in the middle. And history of industry and commerce.
To bring more math and logic into history - one can prioritize timelines too.
And think globally over locally. Prioritize what over who.
----------------------------
Result - the history of math and science can actually help with the study of math and science.
Know how an equation is derived rather than memorizing it.
Know when it was known about. What was before. What was after.
Then it all makes much better sense.
Years ago I coached a girl here who was struggling with her math exams.
She wanted to be a nurse but knew she wouldn't make it if the math within biology was too much for her.
I explained exactly how to study and prepare.
How to make progressively neater handwritten notes in a jotter. On her own.
Make sure she knew how an equation was derived instead of memorizing.
Math became easy for her and she went on with her career.

MEGACHE3SE

its absolutely wild how tygxc is now rejecting even the concept of bayesian probability, when not even a month ago he was swearing by the claim that there was a gradient of mathematic truth, and was making random probabilities for the truth of his claims that the ICCF games are perfect.

playerafar
MEGACHE3SE wrote:

its absolutely wild how tygxc is now rejecting even the concept of bayesian probability, when not even a month ago he was swearing by the claim that there was a gradient of mathematic truth, and was making random probabilities for the truth of his claims that the ICCF games are perfect.

He doesn't have 'the other guy' running interference for him anymore - making T 'look good' in comparison.
Plus moderator entry.
Same thing in the 'hoax' forum. With the same 'other guy' no longer running interference for the climate skeptics.
And again - moderator entry when whoever tried to 'substitute' for the O-person.
Result: again the real issues better exposed and discussed.

MEGACHE3SE

yeah optimissed getting perma muted definitely made laughing at tygxc a lot more streamlined

playerafar
MEGACHE3SE wrote:

yeah optimissed getting perma muted definitely made laughing at tygxc a lot easier

We don't know if O was 'perma' dealt with though.
My guess: he's furious with the mods (while he should be blaming himself which will never happen) ... 
so in PM's O maybe gets himself 'extended'. My guess.
Extended vacations.
And yes 'laughing' at T - 
who 'knows better than ...'
Right.
Difference between T and O ...
T knows that forums like this are 'privileges' and is therefore careful not to abuse the privilege. No 'power plays' from T.
He's got the intelligence there.

MEGACHE3SE
playerafar wrote:
MEGACHE3SE wrote:

yeah optimissed getting perma muted definitely made laughing at tygxc a lot easier

We don't know if O was 'perma' dealt with.
My guess: he's furious with the mods (while he should be blaming himself which will never happen) ... 
so in PM's O maybe gets himself 'extended'. My guess.
Extended vacations.

i dont think so, O was the type to complain about literally EVERYTHING.

Cheese

@Deadpool

MEGACHE3SE
tygxc wrote:

Likewise you cannot say the Riemann Hypothesis is 99% true. It is either true or false.
Schrödinger's cat can be 50% dead, 50% alive, because you can open 100 boxes with Schrödinger cats and find 50 dead and 50 alive.

tygxc's complete misunderstanding of probability aside, I do want to point out that this is literally not how schrodingers cat works.

playerafar
MEGACHE3SE wrote:
playerafar wrote:
MEGACHE3SE wrote:

yeah optimissed getting perma muted definitely made laughing at tygxc a lot easier

We don't know if O was 'perma' dealt with.
My guess: he's furious with the mods (while he should be blaming himself which will never happen) ... 
so in PM's O maybe gets himself 'extended'. My guess.
Extended vacations.

i dont think so, O was the type to complain about literally EVERYTHING.

Yes. But O 'on extended vacations'. Vacations plural.

playerafar
MEGACHE3SE wrote:
tygxc wrote:

Likewise you cannot say the Riemann Hypothesis is 99% true. It is either true or false.
Schrödinger's cat can be 50% dead, 50% alive, because you can open 100 boxes with Schrödinger cats and find 50 dead and 50 alive.

tygxc's complete misunderstanding of probability aside, I do want to point out that this is literally not how schrodingers cat works.

Yes. MEGA well informed.
T twists Everything but then the informed people here Untwist it.
Lol!
Do such things go on in schools?
Up to a point yes.
In various kinds of exams - True or False questions are asked.
Speeds the exams up and means more questions can be put to the students.
Multiple Choice is a variation of the same thing.
Then there's 'solve this'
And 'short essay' answers required.
I encountered large quantities of all those categories of exam questions throughout school.
------------------------------------
T seems to be in charge of the 'false' options.
And Elroch and Dio and MEGA and BC and mpaetz and MarAttigan and llama and others in charge of the True options.
The forum is a kind of Exam.

tygxc

@11326

"Bayesian probability - the only type relevant here,
and not the Frequentist paradigm to which you refer"

++ I appreciate your lecture on Bayesian probability and I note your rejection of the frequentist paradigm, but what in your opinion is the probability that chess is a draw?

To me it is 100%. To me +1 tempo = +1/3 pawn < 1 pawn, not enough to win,
and the dilution by each further move as two compelling arguments.
The ootcome of the 116 perfect games confirms it.

Now what is your Bayesian probability that chess is a draw?

playerafar

True or False:
'Its known that chess is a draw with best play by both sides.'
Correct answer: False.
Can't be known either way. Insufficient information at this time.
--------------------------------------------------
True or False: 'Best play by both sides happens when the strongest chess engines draw each other.'
Correct answer: False.
Can't be known at this time if best play for both sides 'has happened' for multiple reasons as follows.
1) chess isn't solved. So 'perfect play' isn't known.
2) complex issues as to how engines are programmed to 'play for a win' or 'play for a draw'.
3) complex issues concerning as to the 'horizons' of the strongest engines being similiar and therefore resulting in inability to punish imperfect play.
4) complex issues concerning the amount of time alotted to each engine to make each move.
Conclusion: Strong engines drawing each other proves nothing.

MaetsNori
tygxc wrote:

To me it is 100%. To me +1 tempo = +1/3 pawn < 1 pawn, not enough to win

1 tempo is certainly enough to win, in given situations.

White wins by 1 tempo - a example of a hypothetical conversion of White's first-move advantage into a win.

Black is also one move away from delivering his own checkmate - but is one tempo too late.

This is why I (and I'm sure others) have suggested that to ensure that chess is a draw (logically speaking), Black should be allowed to respond to checkmate with his own checkmate on the very next half-move, thus regaining equality and attaining a draw yet again.

Since chess doesn't have this rule, though, it leaves open the possibility that there may be a line somewhere like the position above - where White's 1-tempo advantage proves decisive, despite Black's best attempts to avoid it.

Such a line might not exist, of course - but if we're aiming for rigor, all possibilities need to be considered ...

MEGACHE3SE
tygxc wrote:

@11326

"Bayesian probability - the only type relevant here,
and not the Frequentist paradigm to which you refer"

++ I appreciate your lecture on Bayesian probability and I note your rejection of the frequentist paradigm, but what in your opinion is the probability that chess is a draw?

To me it is 100%. To me +1 tempo = +1/3 pawn < 1 pawn, not enough to win,
and the dilution by each further move as two compelling arguments.
The ootcome of the 116 perfect games confirms it.

Now what is your Bayesian probability that chess is a draw?

100% means your claims are deductively proven.

they quite obviously arent.

its not even a bayesian probability you're just making stuff up.

MEGACHE3SE
MaetsNori wrote:

Such a line might not exist, of course - but if we're aiming for rigor, all possibilities need to be considered ...

this is what tygxc fundamentally refuses to acknowledge, and seems to struggle to understand. he thinks he can just dismiss seemingly unlikely things at his own personal convenience.

the examples you gave and more have all been pointed out to tygxc dozens of times, he just ignores them.

Elroch

1 tempo can also be enough to lose. wink.png

playerafar
Elroch wrote:

1 tempo can also be enough to lose.

Elroch I guess you know the term there.
Concerning that pawn position.
Trebuchet. I believe so.
It sounds like Sean Connery saying 'Treb you Shay' without being drunk.

Elroch
tygxc wrote:

@11326

"Bayesian probability - the only type relevant here,
and not the Frequentist paradigm to which you refer"

++ I appreciate your lecture on Bayesian probability and I note your rejection of the frequentist paradigm

No. We BOTH reject the relevance of the frequentist paradigm in this context. I then informed you that the Bayesian paradigm applies.

, but what in your opinion is the probability that chess is a draw?

While it is clear that neither 0 nor 1 is appropriate, the lack of any disciplined way to estimate the probability means the answer is necessarily subjective.

There are many events where quantitative estimation of a probability is unclear. For example, if I asked you what is the probability that you will be struck by a falling plane in the next year, you would likely accept that the probability is non-zero, but would be unable to give a precise number without doing some research and coming up with a model (which would be non-unique and have some subjective choices). That is the reality of dealing with a stochastic world! We strive to do the best we can.

To me it is 100%. To me +1 tempo = +1/3 pawn < 1 pawn, not enough to win,
and the dilution by each further move as two compelling arguments.
The ootcome of the 116 perfect games confirms it.

None of that makes any sense, because no element of it could convert an uncertain belief into a certain belief in someone who was able to reason correctly.

Now what is your Bayesian probability that chess is a draw?

It is 1-delta, where delta is some small positive number that I don't have a good method to be more precise about. If all you require is an estimate that is better than yours, I pick

1-10^(-10^6) for technical reasons. I am overshooting because it is the safest way for my estimate to be better.

You state that you have arrived at a position of certainty about the result of chess. It is certainly the case that at some time you were not certain of this. Logically speaking there was a single piece of information that first tipped you from uncertainty to certainty. What was it?

This is a rhetorical question, making clear that you do NOT have justifiable certainty, because no plausible answer to the question would justify your belief. You will ignore the question because trying to answer it would refute your erroneous position. Note that by contrast, I did answer your question, and explained that there is subjectivity in the answer (just not in whether it can be 1).