@8722
"If you randomly draw 106 white balls" ++ An ICCF WC Finals Game of 2 years is no ball drawn from an urn, no MNM from a jar, and no coin tossed.
"the urn contains no black balls"
++ I predict nothing about future games, I observe the 106 finished games and conclude the error distribution should be 106 - 0 - 0 - 0 - 0. Even if it were 105 - 0 - 1 - 0 - 0, we still have 105 perfect games and at least a subset of the weak solution of Chess because of the redundancy.
@8722
"If you randomly draw 106 white balls" ++ An ICCF WC Finals Game of 2 years is no ball drawn from an urn, no MNM from a jar, and no coin tossed.
"the urn contains no black balls"
++ I predict nothing about future games, I observe the 106 finished games and conclude the error distribution should be 106 - 0 - 0 - 0 - 0. Even if it were 105 - 0 - 1 - 0 - 0, we still have 105 perfect games and at least a subset of the weak solution of Chess because of the redundancy.
But your method is discredited. Just look back a few posts. (Or a lot of posts if you have the patience.)
And predicting the results of ICCF games has no more relevance to the topic than tossing coins.
@tygxc, you have indicating not understanding the size of a weak solution of chess. That tournament wouldn't scratch the surface. It probably does not even have a single game starting with 16 of the legal white first moves, so does not contain a weak solution against those moves.
Just to check if you are capable of any progress, do you understand that there are two unconnected use of the word "induction" and you have been confusing them?
@8724
"2 games against finalists who made bad decisions"
++ You should really play an ICCF WC qualifier, then you will know.
@8721
"Stockfish was handicapped by its jockeys"
++ The Paris Olympics should ban jockeys from equestrian sports and let the horses run on themselves.
@8722
"If you randomly draw 106 white balls" ++ An ICCF WC Finals Game of 2 years is no ball drawn from an urn, no MNM from a jar, and no coin tossed.
This reveals your lack of understanding of the nature of randomness in EMPIRICAL data. If you had the required cognitive skills, I could explain this to you. I will try to do so, but am pessimistic about and advance in understanding.
The "urn" is a hypothetical enormous set of games just like those in the ICCF world championship. They involve the same players, the same engines, the same procedures, the same randomness by design in the search algorithms of the engines (I understand Stockfish' search algorithm is partially random, so it doesn't always reach the same choices).
The 106 games are a random sample of the games in this urn. They are all draws but they do NOT tell us that the results of ALL of the games in the urn are draws. The results indicate that it is UNLIKELY that a LARGE proportion of the results are wins. The results are entirely consistent with a SMALL percentage (say a fraction of 1%) of the games being wins.
As I say, I am not optimistic, but is that so hard?
[It's really just the same as the variation in your own results, just more boring statistics. You probably win around a half of your games, but there is uncertainty in every individual result until the game has been played. The result depends on the moves played and has no randomness given those moves, but you don't know what they will be before the game].
@8727
"It probably does not even have a single game starting with 16 of the legal white first moves"
++ They only have 1 d4, 1 e4, and 1 Nf3, which are the 3 best moves. The other moves cannot be better from logical reasoning. That is chess knowledge acquired by AlphaZero with no human input but the Laws of Chess, i.e. deduced from axioms. That is Best-first search. If black can draw against 1 e4, 1 d4, 1 Nf3 and 1 c4, then a fortiori it is trivial to draw the other moves.
You are confusing Mathematical induction and Inductive reasoning, not me.
@8727
"It probably does not even have a single game starting with 16 of the legal white first moves"
They only have 1 d4, 1 e4, and 1 Nf3, which are the 3 best moves. The other moves cannot be better from logical reasoning.
You consistently indicate not having a clue what logical reasoning is.
That is
uncertain
chess knowledge acquired by AlphaZero
by inductive reasoning
with no human input but the Laws of Chess, i.e. deduced from axioms. That is Best-first search. If black can draw against 1 e4, 1 d4, 1 Nf3 and 1 c4, then a fortiori it is trivial to draw the other moves.
That is an inductive conclusion. i.e. uncertain. And you ascribe certainty to an engine that is weaker than the strongest?
You are confusing Mathematical induction and Inductive reasoning, not me.
Stupid lies don't help.
Mathematical induction plays no role in this discussion and inductive reasoning NEVER PROVES ANYTHING. You incorrectly referred to "inductive proofs" in post #8712 - you can't deny it.
If arrogance was understanding, you'd be a genius. Your problem is that you permanently think like someone playing blitz chess, confidently attacking, blundering, being certain, being wrong and moving on to the next game.
That is not inadequate for this discussion, or any other one about objective truth.
@tygxc, do you now understand that 106 results in the ICCF world championship are a random sample from the population of all such possible games between the same players?
A simple "no" will suffice. Or silence, indicating "no".
@8727
"It probably does not even have a single game starting with 16 of the legal white first moves"
++ They only have 1 d4, 1 e4, and 1 Nf3, which are the 3 best moves. (by big red telephone) The other moves cannot be better from logical reasoning. (logical reasoning=big red telephone) That is chess knowledge acquired by AlphaZero with no human input but the Laws of Chess, i.e. deduced from axioms. (Not deduced from the axioms; a statistical result from a number of games played with AZs limited look ahead.) That is Best-first search. (What is Best-first search? Best to read your links before posting them.) If black can draw against 1 e4, 1 d4, 1 Nf3 and 1 c4, then a fortiori it is trivial to draw the other moves. (The logic is, of course, profoundly stupid, but since you believe Black can draw against those moves why don't you post us examples of you trivially drawing against Stockfish from the other moves. We don't have anything perfect for you to play, but against SF would go some way to raising your credibility.)
You are confusing Mathematical induction and Inductive reasoning, not me.
I don't believe he is. Whatever it is you're doing, it doesn't give the right results.
@8731
"If arrogance was understanding, you'd be a genius." ++ That applies to yourself.
I will try to explain again.
We have 106 draws out of 106 games. I say the distribution of errors is 106 - 0 - 0 - 0 - 0.
You say e.g. 1 e4 wins for white.
I say no I have 2 games 1 e4 e6 that draw.
You say, no 1...e6 is a mistake and after 2 d4 d5 they played 3 Nc3 which is a mistake, as 3 e5 wins for white.
I say, OK, let us admit that, 2 games with 2 errors each, error distribution 104-0-2-0-0.
I say I have 15 games with the Sicilian 1 e4 c5 that draw.
You say, no 1...c5 is a mistake and they played 3 Nf3 and 2 Nc3, when in fact 2 c3 wins for white.
I say OK, let us admit that, 15 games with 2 errors each, error distribution 89-0-17-0-0
I say I have 6 games with the Petrov 1 e4 e5 2 Nf3 Nf6 that draw.
You say, no 2...Nf6 is a mistake, when in fact 3 Nc3 wins for white.
I say OK, let us admit that, 6 games with 2 errors each, error distribution 83-0-23-0-0.
I say I have 14 games with 1 e4 e5 2 Nf3 Nc6 that draw.
You say, no 2...Nc6 does not help either and in fact 3 Nc3 wins for white.
I say OK, let us admit that, 14 games with 2 errors each, error distribution 69-0-37-0-0.
However, this is no longer plausible. Why would there be 69 games with 0 error and 37 with 2 errors and none with 1 or 3 errors? Why would all errors be paired?
Moreover this level of errors is not possible as humans at 3 min/move arrive at 1.1 error average per game.
@8724
"2 games against finalists who made bad decisions"
++ You should really play an ICCF WC qualifier, then you will know.
@8721
"Stockfish was handicapped by its jockeys"
++ The Paris Olympics should ban jockeys from equestrian sports and let the horses run on themselves.
Horses can run faster without jockeys. They often run along just for fun after unseating their jockeys, while the jockeys need to encourage their steeds to maintain the same speed.
Jockeys add weight. The exact weight matters enough for handicapping to be used, where a jockey carries a little extra weight (a few pounds) to compensate based on past results.
@8731
"If arrogance was understanding, you'd be a genius." ++ That applies to yourself.
I will try to explain again.
We have 106 draws out of 106 games. I say the distribution of errors is 106 - 0 - 0 - 0 - 0.
You say e.g. 1 e4 wins for white.
Fake quote.
I say no I have 2 games 1 e4 e6 that draw.
You say, no 1...e6 is a mistake and after 2 d4 d5 they played 3 Nc3 which is a mistake, as 3 e5 wins for white.
Fake quote.
I say, OK, let us admit that, 2 games with 2 errors each, error distribution 104-0-2-0-0.
I say I have 15 games with the Sicilian 1 e4 c5 that draw.
You say, no 1...c5 is a mistake and they played 3 Nf3 and 2 Nc3, when in fact 2 c3 wins for white.
Fake quote.
I say OK, let us admit that, 15 games with 2 errors each, error distribution 89-0-17-0-0
I say I have 6 games with the Petrov 1 e4 e5 2 Nf3 Nf6 that draw.
You say, no 2...Nf6 is a mistake, when in fact 3 Nc3 wins for white.
Fake quote.
I say OK, let us admit that, 6 games with 2 errors each, error distribution 83-0-23-0-0.
I say I have 14 games with 1 e4 e5 2 Nf3 Nc6 that draw.
You say, no 2...Nc6 does not help either and in fact 3 Nc3 wins for white.
Fake quote.
I say OK, let us admit that, 14 games with 2 errors each, error distribution 69-0-37-0-0.
However, this is no longer plausible. Why would there be 69 games with 0 error and 37 with 2 errors and none with 1 or 3 errors? Why would all errors be paired?
Moreover this level of errors is not possible as humans at 3 min/move arrive at 1.1 error average per game.
How do you think lying about what I say five times advances this discussion? Discuss what I say, like I discuss what you say (including quotes).
Big red telephone again?
Humans are probably better than Stockfish then. Stockfish can manage 13 blunders after the game's already down to 7 men.
Are you sure you didn't miss some blunders when you counted them? It's easy to do if you haven't got a tablebase.
""these engines are just reallly strong"
++ ICCF (grand)master + engines at average 5 days/move
are much stronger than engines running unjockeyed,
and these are much stronger than the strongest humans at 3 minutes/move."
so what you are telling me is "these engines are just reallly strong"
thats all.
I agree. And you caught tygxc in at least one contradiction and I think I have too.
And - again - he has already conceded chess can't be solved.
@tygxc, you have indicating not understanding the size of a weak solution of chess. That tournament wouldn't scratch the surface. It probably does not even have a single game starting with 16 of the legal white first moves, so does not contain a weak solution against those moves.
Just to check if you are capable of any progress, do you understand that there are two unconnected use of the word "induction" and you have been confusing them?
there it is. Terminology.
One of the bear traps for those who refuse or otherwise fail to grasp the generic nature of science.
@8664
"the mistakes are smaller"
++ There are no smaller mistakes. A mistake (?) is a move that either changes a drawn position to a losing position, or a won position back to a drawn position a.k.a. missed win.
A blunder or double mistake (??) is a move that changes a won position to a lost position.
Wrong.
Mistakes can be smaller or larger.
And are.
A mistake that drops a queen will so often be game-deciding.
A mistake that drops a pawn will often not be game-deciding.
Of course mistakes aren't binary.
But tygxc is like a determined defense lawyer.
But there's a 'rub'.
He's like his own jury - he 'plays' that jury.
Since he's his own jury - then in his mind - he always wins.
But its inoccuous I'd say.
Not grandiose. Not psychotic. Not even personality disorder level.
He just disagrees. Proves he can do so without trolling.
---------------------------------
Whereas the muted guy ... totally different situation. The O-person.
@tygxc is correct from the point of view of an oracle, like a 32 piece tablebase. The problem is that we don't have access to an oracle, and he thinks a feeble approximation must be good enough without any good reason.
Remember that having 106 draws is not even good evidence we won't see a win in the next 106 games.
It's worth observing that the ICCF games give a very skewed view because they always compare the current best engine to itself. If instead you compare the 2024 best engine to the 2019 best engine it will win some games - it has a higher rating. Likewise, it is likely that if you play the 2029 best engine against the 2024 best engine, it will win some games, despite the 2024 best engine getting 106 draws against itself.
Stockfish has improved by over 800 points in the last 10 years
Yes - Elroch put it very well there. Especially in the last two paragraphs.
He has the details.
Oh my gosh! I made three posts in a row! Or rather column.
Broke my policy.
Egads!
-------------------------------
And Elroch said something like I was saying - but he said it better.
"Likewise, it is likely that if you play the 2029 best engine against the 2024 best engine, it will win some games, despite the 2024 best engine getting 106 draws against itself."
Yes! The key!
and 800 points? That's Huge.
Even just half of that - that means the engine's not just winning games against the old weaker engine -
its winning a gigantic percentage of games!
The weaker engine would be struggling to get even one draw out of every ten games ...
@8692
"And every theorem in the mathematical sciences (including computer science and game theory) has NO uncertainty."
++ Many have, like Shannon's theorem, the Uncertainty principle and many more.
oh more lack of knowledge in your end. that's an entirely different type of uncertainty.
its pretty clear that you only see the words and make up your own understanding of them, and then when you see the same words in different contexts you automatically assume its your own context.
@8731
"If arrogance was understanding, you'd be a genius." ++ That applies to yourself.
I will try to explain again.
We have 106 draws out of 106 games. I say the distribution of errors is 106 - 0 - 0 - 0 - 0.
You say e.g. 1 e4 wins for white.
I say no I have 2 games 1 e4 e6 that draw.
You say, no 1...e6 is a mistake and after 2 d4 d5 they played 3 Nc3 which is a mistake, as 3 e5 wins for white.
I say, OK, let us admit that, 2 games with 2 errors each, error distribution 104-0-2-0-0.
I say I have 15 games with the Sicilian 1 e4 c5 that draw.
You say, no 1...c5 is a mistake and they played 3 Nf3 and 2 Nc3, when in fact 2 c3 wins for white.
I say OK, let us admit that, 15 games with 2 errors each, error distribution 89-0-17-0-0
I say I have 6 games with the Petrov 1 e4 e5 2 Nf3 Nf6 that draw.
You say, no 2...Nf6 is a mistake, when in fact 3 Nc3 wins for white.
I say OK, let us admit that, 6 games with 2 errors each, error distribution 83-0-23-0-0.
I say I have 14 games with 1 e4 e5 2 Nf3 Nc6 that draw.
You say, no 2...Nc6 does not help either and in fact 3 Nc3 wins for white.
I say OK, let us admit that, 14 games with 2 errors each, error distribution 69-0-37-0-0.
However, this is no longer plausible. Why would there be 69 games with 0 error and 37 with 2 errors and none with 1 or 3 errors? Why would all errors be paired?
Moreover this level of errors is not possible as humans at 3 min/move arrive at 1.1 error average per game.
every single move in all of those games are potential errors, you have no way of proving otherwise
the fact that you narrowed the selection down to just the opening is stupidity or intellectual dishonesty on your end.
@8727
"It probably does not even have a single game starting with 16 of the legal white first moves"
++ They only have 1 d4, 1 e4, and 1 Nf3, which are the 3 best moves. The other moves cannot be better from logical reasoning. That is chess knowledge acquired by AlphaZero with no human input but the Laws of Chess, i.e. deduced from axioms. That is Best-first search. If black can draw against 1 e4, 1 d4, 1 Nf3 and 1 c4, then a fortiori it is trivial to draw the other moves.
You are confusing Mathematical induction and Inductive reasoning, not me.
LOL now you are just moving the goalposts to keep your fantasy alive, while still getting the basic definitions wrong. its pretty clear that you dont actually know what's in these articles you cite, considering how the last two times I went in depth with your citations, it turned out you had grossly misrepresented what they were saying.
you still havent addressed how ive shown what you've said to math undergrads, math grads, and math professors, and all of them pointed out the same mistakes that I pointed out to you.
I showed them your posts with your own context givn. what makes you think you know better than all of them?
2 games. LOL.
What evidence do you have that the results were not either random in nature or the result of some players using more powerful hardware than others?
I understand that you have difficulty understanding the word "random", but consider that if you play a long series of games between Stockfish and itself, you will see occasional random wins.