Chess will never be solved, here's why

Define thinking for me please.

Thank you Lola for your continuing persuasiveness (and aplomb). If only there were more of you posting in our "crazy forum threads," then the level of discourse would surely rise.

Any chance you can be cloned? Maybe after you first "walk on Mars?" Best Wishes, Dear.

Alexeivich94

Read dictionary

I'm guessing you need to spend a huge amount of time analyzing, though. It takes an extreme amount of patience to engine-crunch a position for 5 days per move, or however long they devote to each position.

That's one thing I'd tip my hat to ICCF players over: playing a single game for months or even years isn't something I'd be capable of tolerating ...

I wonder if anyone is able to tell how strong a recent TCEC game (SF vs L0) might compare to a recent ICCF game, in terms of playing strength? I mean, to me it's all virtually the same, stratosphere-level moves, but ... maybe someone here can spot the differences.

Goo goo Gaga

burst !...lemme guess...u wanna try2make a baby right now right ? lol !

AI is trash. Human brain can only think. AI is a computer that can not think. You cant solve something without understanding it.

lets see. ur lunch was free and ur out to it ?...AI will dominate stuff in 5 to 10 yrs. in fact ur gonna laff at urself then...what u just said (they must be 50+).

4u zborg:

https://www.youtube.com/watch?v=R4gOIt-M02A

...n'da linda rosa lonji id wear lol !

https://www.youtube.com/watch?v=LB-tkHD4uXU

@13241

"you need to spend a huge amount of time analyzing" ++ Yes, even before a game starts:
they analyse all opponent's prior games beforehand looking for positional weaknesses.

"It takes an extreme amount of patience to engine-crunch a position for 5 days per move, or however long they devote to each position." ++ Yes, normally they take 2 to 10 days per move.

"playing a single game for months or even years" ++ It is a huge task.

"how strong a recent TCEC game (SF vs L0) might compare to a recent ICCF game"
++ ICCF is stronger: longer time per move, human intervention for openings, long term planning, positional play. TCEC superfinals have 50 imposed openings, to prevent all draws.

@13233

"I can't imagine I'd lose a single game" ++ Then you should sign up for a WC preliminary.

"a few people do play without an engine" ++ No, not anymore.

"He's rated lower than me OTB"
++ Maybe he lacks the nerves to play over the board with a clock ticking.
Maybe he is good at positional play, but cannot calculate tactics.

"he can't provide the supposed "deep" understanding that humans use to aid engines"
++ He qualified for the WC Finals.

"He played a blunder that my engine immediately recognized as a blunder."
++ It happens. Clerical error, illness, personal problems,
time trouble: 50 days per 10 moves, but some use up all and then have to reply the same day.

"These were (apparently) the best ICCF players of 2020"
++ They qualified: preliminaries, semifinals, candidates, and finally finals.

@13231

"the other wins came from beating Bock, Steffen" ++ He had personal problems.
A game lasts over a year, and things happen in life in a year.

hey tygxc why arent you addressing the fact that i personally had mathematicians verify that your weak solution calculations and methodology were completely delusional?

Thee_Ghostess_Lola If you knew anything about AI you would know that it cant think on your own and is run by human made algorithms. And you are right. It will dominate ignorant people like you...

In their analysis for each move, modern engines leave 95% of the nodes generated by legal moves from positions reached completely unanalysed (typically all but the first or second highest evaluated move). You said "human logic is enough" to ignore these positions.

Could you explain how humans look at the (literally) trillions of unanalyzed nodes in the engine analysis for each individual move, each of which individually has the potential to refute the conclusion of the engine?

And if you are relying on the notion that engines pick the best move all the time, remember that they blunder frequently in 6 and 7 piece tablebase positions when not told what to play by a tablebase.

How reliable do you think they are in 8 piece positions?

dont forget that tygxc only assigns 1 node per position in his calculations.

He understands how the opponent typically has many legal moves, and often several superficially plausible ones.

@13252

"modern engines leave 95% of the nodes generated by legal moves from positions reached completely unanalysed (typically all but the first or second highest evaluated move)"
++ It is the human ICCF player who decides what moves his engine must look into.

"human logic is enough" for some positions. Like 1 e4 e5 2 Ba6?, or 1 a4, or 1 e4 e5 2 Bd3.
It is the human ICCF player who with the help of data bases and engines decides the opening.
One reason for that is that chess it most complicated around 26 men, see Table 3

"how humans look at the (literally) trillions of unanalyzed nodes"
++ The ICCF human player decides what moves to look into.
As ICCF World Champion Ron Langeveld said:
'How many ideas can you interactively throw at the computer in one hour is the key question'

"relying on the notion that engines pick the best move all the time"
++ No not at all. The human ICCF player decides on his move after consulting his engines.

"they blunder frequently in 6 and 7 piece tablebase positions"
++ A) at blitz speed, not 5 days / move, and B) in irrelevant positions.
Weakly solving Chess is hopping from the drawn initial position to other drawn positions to finally reach a certain draw: either a 7-men endgame table base draw, or a prior 3-fold repetition, or an otherwise known drawn position as judged by both finalists and their engines.
This means only drawn positions are relevant.
White wins are pitfalls for black. Black wins are pitfalls for white.

"How reliable do you think they are in 8 piece positions?"
++ I think ICCF player + engines are now 100% reliable in 8-men positions. 9 games ended in a 7-men endgame tablebase draw, so I am confident that they played the 8-men phase reliably.

I am not sure what you mean.

Concerning the number of potentially crucial unanalyzed positions, there is a refinement to my rather casual claim earlier. If a computer manages 300 million nodes per second for 4 days, this is 3.7 x 10^13 nodes. Suppose it is white to move. The branching occurs for both white and black (with a similar branching factor), but loose ends (unanalysed leaf nodes) in branches for candidate branches for white that are not rated #1 can't refute its evaluation of the main line (they are merely additional options it could change to later). This is a 100% reliable type of pruning for the purpose of this calculation.

Then consequence of this is that the number of unanalysed moves that can refute the mainline of analysis is given by around sqrt(3.7 e13) * (true branching factor - analysed branching factor) assuming constant branching factor (the estimate is rather robust to variation of this that is not skewed to one side of the board, which is entirely reasonable).

So, assuming an average true branching factor of 35, this means there are more than sqrt(3.7 e13) * (35 - 2) which is greater than 200,000,000 legal moves in the analysis tree for a single move at an undecided stage of an ICCF game. Any of these could refute the assessment of the move played (i.e. leave black with a win if it is white to play).

So, just a few little holes to fill to solve chess this way. Well, about 200,000,000 per move of any ICCF games you want to rely on.

[Warning - fixing each individual hole properly would generally be a task beyond practical computation. It's solving a game that could be almost as hard as chess]