@4024
"In any case, the numbers cannot be achieved in 5 years."
++ Because you say so? I take the word of Sveshnikov over yours. The numbers confirm it.
"I believe that you give him too much credit."
++ I believe you do not pay him enough respect. He knew what he was talking about.
"Trillions of positions in five years?" ++ Yes: a billion positions per second per cloud engine.
"Each with human intervention?" ++ No. Human intervention only at the start and at the end.
@4024
"In any case, the numbers cannot be achieved in 5 years."
++ Because you say so? I take the word of Sveshnikov over yours. The numbers confirm it.
"I believe that you give him too much credit."
++ I believe you do not pay him enough respect. He knew what he was talking about.
"Trillions of positions in five years?" ++ Yes: a billion positions per second per cloud engine.
"Each with human intervention?" ++ No. Human intervention only at the start and at the end.
I'm just curious, when does the 5 year period start? Was it a year ago? 4 years ago? I know it's not today.
@4028
"when does the 5 year period start?"
++ When there is money for the cloud engines and the assistants.
@4030
"you clearly state that each positional assessment is guided by a human GM"
++ No, not at all. The humans start the calculations from meaningful positions.
Not 1 e4 e5 2 Ba6, not 1 a4, not 1 Nf3 d5 2 Ng1.
The humans also select the promising lines, e.g. 1 e4 e5, renouncing all other 19 replies.
The humans stop a calculation when it is a clear draw or win. E.g. when one side is material down with no compensation or in some opposite colored bishop endings. That is to save engine time. In case of any doubt, calculate more.
"It's impossible for humans to deal with such numbers." ++ The humans do not deal with the numbers, the engines do. The humans only reduce the number the engines have to deal with.
"he would have thought it possible" ++ Sveshnikov clearly stated so in his farewell interview.
"why YOU think it possible" ++ Because I have calculated the number of legal, sensible, reachable during solving, and relevant positions to be 10^17 and that corresponds to 5 years.
"Trillions of positions in five years?"
++ Yes: 5 a * 365.25 d/a * 24 h/d * 3600 s/h * 10^9 position/s/engine = 10^17 position/engine
"Each with human intervention?" ++ No, not at all. The humans only start the calculations and occasionally end a calculation in case of clear win or draw.
"why you can think it possible that a human can personally intervene"
++ The humans set up the starting positions.
The humans occasionally end a calculation that has arrived at a clear draw or win.
"Trillions of positions in five years?" ++ Yes 10^17 positions.
"Each with human intervention?" ++ No. The humans start the calculations.
The humans occasionally end a calculation whenever they notice a clear draw or loss.
@4031
I give a few examples, all ICCF WC drawn games, 99% sure to be perfect games.
https://www.iccf.com/game?id=1164344
This ends in a 7-men endgame table base draw, no human adjudication required.
https://www.iccf.com/game?id=1164259
This needs human adjudication otherwise the engine must burn useless time in this clear draw.
https://www.iccf.com/game?id=1164319
This ends in a perpetual check, i.e. a forced 3-fold repetition. No human adjudication needed.
@4033
"We have already established that the human intervention is necessary to determine which positions are "meaningful" and need guidance"
++ No, we have not. The humans launch the calculations from meaningful starting positions.
The humans occasionally end calculations in clearly drawn or won positions.
"there is no algorithm that is 100% reliable where the positional assessment is difficult"
++ Positional assessment is calculation. A positional advantage is an advantage that transforms into a material advantage or ultimately checkmate in a certain number of moves.
The humans make no positional assessments. The humans launch the calculations.
The humans end calculations in case of a clear draw or win.
"You continue to evade the issue."
++ I do not evade any issue. You do not understand and you twist my words.
Try to read and understand. It is not that hard.
@4028
"when does the 5 year period start?"
++ When there is money for the cloud engines and the assistants.
Ah. How convenient. I thought when you said chess would be solved in 5 years you meant from now, or when you first made the claim. I didn't know the 5 year period starts at some random undetermined time in the future.
My teenage nephew said he is going to be a millionaire in 2 years. Now I know what he meant.
So out of curiosity why 5 years? If the clock doesn't start until some point in the future none of us know, why not say 1 year? Or even one month, or 5 days?
@4035
"tygxc is not endeavouring to arrive at an honest assessment" ++ Yes, I am
"merely to sell a story" ++ I am not selling anything. I am not even applying for an R&D grant.
"where he is clearly either innocently wrong" ++ Nowhere
"or attempting to deceive." ++ Nowhere. You on the other hand twist my words.
@4036
"you are constantly backtracking and changing your story" ++ No, I say the same thing.
"the impression your argument gives is that you are being dishonest"
++ You do not read, you do not understand and that gives you an erroneous impression.
"There is absolutely no question that you have now lost your argument"
++ There is absolutely no question that no valid couterargument has been presented.
"your answers are insufficient to convince anyone"
++ If you do not read or understand, I cannot help.
I patiently explain, answer questions, quote literature, give examples.
"there are trillions of positions to be considered, in five years" ++ Yes, 10^17.
"You have admitted that the present algorithms in Stockfish are insufficient"
++ I have calculated that there is < 1 error in 10^20 positions that the objectively correct move is not within the top 4 moves of Stockfish when running for 17 s on a 10^9 nodes/s engine, or for 17000 s on a 10^6 nodes/s desktop. There are no 10^20 positions, only 10^17 positions.
"You have accepted that a large number of human interventions would be necessary."
++ I estimate 3 grandmasters full time 40 h/week during 5 years.
"Trillions of positions must be assessed." ++ Yes, 10^17
"Millions of positions require human input."
++ No. The proof tree is expected to contain 1 billion positions, that is 10 million games.
"If one GM can consider 1000 positions per day"
++ Kramnik said he studied 10,000 games per month.
3 grandmasters * 10,000 games/month/grandmaster * 12 month/a * 5 a = 1.8 million games. Most of the games need no intervention, see examples @4033.
So intervention in 18% of games seems plausible.
"You claim that is going to be 100% accurate is a disconnect with reality."
++ Losing Chess and Checkers also got 100% accurate proofs.
"Yes, because I say so and many others also say so." ++ So if 10 players of 1700 strength say one grandmaster does not know chess analysis, then the majority is right?
"It is clear that you do not respect the word of others you're arguing with."
++ I have so far not seen one valid counterargument.
"should they respect your word?"
++ They should, but they do not, I get insults, often a sign of lacking argument.
Anyway nobody should accept my word, I am no Sveshnikov, I do not even have a master title. People should try to read and understand before dismissing without any argument.
"Everything you have written is grotesquely self-serving" ++ I do not benefit from it.
"circular" ++ no
"based on your fantasy" ++ Based on facts & figures.
"You should be honest." ++ I am. You should be too.
"Otherwise you will get no respect."
++ I get no respect from the disrespectful. I am not after respect, I am after the truth.
"You've lost your argument"
++ Chess is about winning, or losing, or drawing, an argument is about the truth.
@4037
I thought when you said chess would be solved in 5 years you meant from now.
++ Humans can walk on Mars within 10 years after allocation of resources.
"So out of curiosity why 5 years?"
++ That is what GM Sveshnikov said and that is what the facts & figures confirm.
@4037
I thought when you said chess would be solved in 5 years you meant from now.
++ Humans can walk on Mars within 10 years after allocation of resources.
"So out of curiosity why 5 years?"
++ That is what GM Sveshnikov said and that is what the facts & figures confirm.
But none of that makes any sense. Why 5 years if the starting point is some unknown future time? We could also say humans can walk on mars within 10 minutes. If the starting point is after the spaceship has already landed there.
So it sounds like this Sveshnikov character is saying chess will be solved in 5 years. Starting in an unknown number of centuries from now when computers will be strong enough to do it.
@4033
"tygxc, you're accusing me of twisting your words" ++ Yes, you did that.
"You aren't "after the truth"." ++ Yes I am.
"You are not willing to take good arguments into consideration."
++ I have not read any good counterargument so far. Lots of insults, ridicule, ad hominem...
"You merely repeat your incorrect ones" ++ I repeat the correct ones.
"It doesn't work." ++ People do not read, do not understand, though claiming to be intelligent.
"You've lost your argument." ++ I argue for the truth, not for winning or losing unlike chess.
"How is your Alzheimer's Disease?" ++ Speak for yourself.
@4034
"Why 5 years if the starting point is some unknown future time?"
++ How long does it take to build a house? When does that period start?
The starting point is when the money is there.
The work cannot start before the assistants and computers are there to start it.
"So it sounds like this Sveshnikov character is saying chess will be solved in 5 years."
++ Read again:
'Give me five years, good assistants and the latest computers
- I will bring all openings to technical endgames and "close" chess.' - GM Sveshnikov (+)
The 'give me' expresses the necessity of the assistants and computers as a prerequisite.
The 'good assistants' imply human assistance being essential. He even names them first.
The 'lastest computers' implies that they already exist.
You do not seem to know him
https://en.wikipedia.org/wiki/Evgeny_Sveshnikov
https://www.chess.com/news/view/evgeny-sveshnikov-1950-2021
@4036
"So the number of seconds in five years, divided by your figure of 17 seconds per position is a figure somewhat less than 10 millions."
++ In 17 seconds the engine calculates 17 billion positions.
You confuse the proof tree and the search tree.
The engine considers far more positions than it retains.
For comparison: Checkers has a proof tree of 10^7 positions and a search tree of 10^14.
"++ No, we have not. The humans launch the calculations from meaningful starting positions.
The humans occasionally end calculations in clearly drawn or won positions."
Are the meaningful starting positions determined by engine evaluation? What range of evaluations are considered the relevant positions (ie. -1 to +1?) Isn't the margin of error for current engines too big to reliably confirm the relevant positions?
@4042
"Are the meaningful starting positions determined by engine evaluation?"
++ No, they are not: the human assistants determine them with the help of data bases.
"Isn't the margin of error for current engines too big to reliably confirm the relevant positions?"
++ Yes, and that is why the engine evaluation is not used for determining the meaningful starting positions. Chess is most complicated at 26 men: Table 3.
https://arxiv.org/pdf/2112.09386.pdf
So the human assistants with data bases select the meaningful positions of 26 men.
From the initial position 1 e4, 1 d4, 1 c4, 1 Nf3 are most meaningful and have to be handled.
1 a4 is not meaningful, it does not try to win. That is based on human logic, but has been confirmed by independent engine calculations (no evaluation, but whole games) with no other input but the Laws of Chess: Figure 31.
https://arxiv.org/pdf/2111.09259.pdf
Looking at 1 e4 the humans select a promising reply to draw, say 1...e5. (They could also select 1...c5.) Now there are several white tries to win: 2 Nf3, 2 Bc4, 2 Nc3, 2 d4 to treat.
2 Ba6 needs no consideration: we know it loses, so it does not try to win.
Some of the moves transpose, the human assistants take care not to duplicate.
E.g. 1 e4 e5 2 Nf3 Nc6 3 Bc4 = 1 e4 e5 2 Bc4 Nc6 3 Nf3
or 1 e4 e5 2 d4 exd4 3 Nf3 Nc6 = 1 e4 e5 2 Nf3 Nc6 3 d4 exd4
Looking at 2 Nf3 the humans select a promising reply to draw, say 2...Nc6. (They could also select 2...Nf6.) Now there are several white tries to win: 3 Bb5, 3 Bc4, 3 Nc3, 3 d4 to treat.
Looking at 3 Bb5 the humans select a promising reply to draw, say 3...Nf6. (They could also select 3...a6.) Now there are several white tries to win: 4 d4, 4 O-O, 4 d3, 4 Qe2 to treat.
It goes on like that and the good asssistants arrive at positions like this one:
This has 26 men and here engine evaluations are getting accurate enough to launch the engine. Each time the top 4 white moves are investigated and the top 1 black reply is chosen after thinking 17 seconds on a cloud engine, which corresponds to 4.7 hours on a desktop.
As calculated the objectively correct move will always be among the top 4 engine moves if the cloud engine runs for 17 seconds, corresponding to 4.7 hours on a desktop, with only 1 error in 10^20 positions, but there will be only 10^17 positions.
It is even more efficient to start from an ICCF WC drawn game, as calculations show it already is 99% sure to be a perfect game with no error and as it represents 2 years of engine and grandmaster effort. Example:
https://www.iccf.com/game?id=1164344
This game ends in a 7-men endgame table base draw.
Look at the last white move 57 Bxa1. What 3 alternatives are there and do they draw too?
If yes, then look at 56 Bb2. Calculate what 3 alternative moves are there and do they draw too?
If yes, then look at 55 Kd3. Calculate what 3 alternative moves are there and do they draw too?
Calculate like that all the way back until a position from another data base game is reached.
Checkers has a proof tree of 10^7 positions [ . . . ]
Emphases mine:
"The stored proof tree is "only" 10⁷ positions. Saving the entire proof tree, from the start of the game so that every line ends in an endgame database position, would require many tens of terabytes, resources that were not available. Instead only the top of the proof tree, the information maintained by the manager, is stored on disk. When a user queries the proof, if the end of a line of play in the proof is reached, then the solver is used to continue the line into the database. This dramatically reduces the storage needs, at the cost of re-computing (roughly two minutes per search)"
https://www.researchgate.net/publication/231216842_Checkers_Is_Solved
@4044
"10^14 is a good ballpark estimate of the forward search effort."
https://www.cs.mcgill.ca/~dprecup/courses/AI/Materials/checkers_is_solved.pdf
Page 1521, column 3, paragraph 3.
"However, human analysts consider this line to be a win for Black, and the
preliminary analysis agreed. Hence, the seven openings beginning with the moves 12-16 and
24-19 were proven instead. This led to the least amount of computing."
Page 1521, column 2, paragraph 2.
"“Are the results correct?” Early on in the computation, we realized that there
were many potential sources of errors, including algorithm bugs and data transmission errors.
Great care has been taken to eliminate any possibility of error by verifying all computation
results and doing consistency checks. As well, some of the computations have been independently verified (SOM text). Even if an error has crept into the calculations, it likely does not change the final result."
Page 1521, Column 1, paragraph 1
"That is _NOT_ a general rule." ++ Pawn up = win is a general rule, but there are exceptions.
So, strictly speaking, it is not a general rule. How many exceptions, in percentage?
Don't be sarcastic.
I was not. @tygxc supported his claim that a pawn of advantage is a win quoting Capablanca. Besides the fact that this is an appeal to authority, Capablanca wrote that "the winning of a pawn among good players of even strength often means the winning of the game" (Chess Fundamentals, emphasis mine), which is rather different from "always wins the game, with exceptions". Engines usually assign the advantage of a pawn an expected score around 0.75.
The reason why I insist on that goes beyond the actual value of a pawn. It is to state that it is not possible, in order to declare chess solved, to cut off lines on the basis of "general" principles not proven to always hold true (in chess, indeed, most of them are proven to not always hold true).
1 e4 e5 2 Ba6 provides no compensation at all and can be dismissed.
Unproven.
A comment like that means that nothing you say can be taken seriously.
That's a pooh-pooh. Stating that 1. e4 e5 2. Ba6 may even win, as above goes beyond the value of the specific line. Allowing dismissals of such lines would create precedents too, and it would be very difficult, then, to fix boundaries for what can or cannot be cut off during the search.
"they can be sound (1. e4 e5 2. f4)" ++ 'It loses by force' - Fischer, 'I could not find a way for white to equalise' - Kramnik '23.4% black wins, 6.3% white wins' - AlphaZero Figure 4. (d)
https://arxiv.org/pdf/2009.04374.pdf
Appeal to authority.
That's completely acceptable when discussing chess lines.
Missing the point. It is not acceptable when discussing about solutions. A solution has to prove whether and when authorities are right, not the other way around.
And after two "ad hominem"
This is becoming repetitive. You ought to have settled for your one good point and not tried to do a bthicker.
Thank you for the tip, but sorry, I am not taking you as an example of good arguing.
The pedantic adherence to using transient and possibly faulty post numbers is directly analogous to the pedantic adherence to using faulty assumptions and numbers in trying to solve chess .
That's something you and I agree on. [ . . . ] there's something going on in this discussion that I don't like the look of. I'm not blaming you because it's happened to us all: getting sucked into a profitless argument that basically doesn't move..
It is not going to move. That's because, despite the inevitable @tygxc's reply, chess cannot be solved in reasonable time, at the moment. But some people never concede, as you yourself noted. So what do you expect? The only thing one can do is to fight attempts of deceiving and disinformation.