Chess will never be solved, here's why

to clarify i deleted because i did not realize the occasion to address you in first person, not third person, which might have seemed rude not to sddres you as

I kept going out and coming back in and in some cases it still preferred my moves and in others it reverted to its own, where you have to play your move, let it think for a bit and when the score goes up, go backwards and then forwards again. So it seemed to have a bit of learning ability, which surprised me. Maybe something stuck in a cache though.

of course it has a learning ability. stockfish is the most powerful chess AI, after all

i gotchu man (sorry if my exprimation sucked before but it is 3 am rn)

Since tygxc downvoted my post without addressing it, Ill add another example where @tygxc tries to make claims of hard fact, (often mathematical), but then when confronted with the fact that he literally has no grounds or definitive logic, he starts speaking in softer terms and acts like its a grey area discussion.

for example, tygxc tried to claim as fact that chess will be strongly solved in 80 years by quantum computers. This requires over 10^44 data points, compared to earth having 10^50 atoms.

turns out, he arrived at this ridiculous claim by just assuming that quantum computers would accelerate at the exponential growth of transistor packing, the single most explosive growth of any technology in all time, while at the same time assuming that quantum computers would have the same quantum advantage with chess operations as other relevant operations, even though that isnt true.

back in reality, nobody in the field has made any such predictions about quantum computing, nor has any quantum program been theorized to extend to chess.

tygxc's admitted personal estimates and feelings were all that support his own claims. But tygxc nevertheless has reposted that 80 year claim as fact on multiple forums.

Only you could think that Stockfish is "learning" from your moves. If you let it sit, it comes up with the best move. Sometimes it takes seconds and sits tight, over times it it takes minutes or hours to change its evaluation. Rest assured, none of that is predicated on your choices, at all.

@14858

"Of course Stockfish can't beat Stockfish - it's limited by its own playing strength. It cannot defeat itself."
++ But player + engines can beat player + engines,
that is what they do in Preliminaries, Semifinals, and Candidates to qualify for the Finals.
That is also what they did in the Finals until this year.

"We've already seen a past ICCF WC game where the draw was a mistake - an error that was easily found by today's technology."
++ Yes in the 2013 ICCF WC Finals there were 20 decisive games and thus expected 2 draws with 2 errors. In the 2016 ICCF WC Finals there were 10 decisive games and thus expected 1 draw with 2 errors. In the ongoing ICCF WC Finals there are 100% draws and thus 114 perfect games with optimal play from both sides and zero error.

"anyone on these forums could travel back in time with a cheap laptop and SF 16.1 installed, enter the 2013 ICCF WC, and dominate the entire field with relative ease."
++ No. That is like saying everybody who can ride a horse would be Olympic jumping champion when given a top horse or everybody who can drive a car can win a formula-1 Grand Prix if given the car of Verstappen. You cannot play ICCF today without engines, but in the end it is the player that makes the difference.

"engines from the year 2034 will be capable of doing the same to today's ICCF field"
++ No, not at 5 days/move. You cannot drop below 0 error/game.
Engines from 2034 will be able to do the same in 5 h/move or 5 minutes/move.

What is going on here?

@14876

"it's easy to understand why errors will come in pairs." ++ Some errors, but not all errors.

"All the competitors are using the best, latest engines and top-flight hardware"
++ The Russians run inferior hardware due to sanctions, but 4 of the 17 finalists are Russian.
In the Preliminaries, Semifinals, and Candidates there are decisive games, that is how the Finalists qualified. The hardware and software is not less equal in those than in the Finals.

ah yes tygxc, please continue to straight up ignore rebuttals/basic facts presented by myself and others in order to keep on pretending your fallacy has weight.

"In the ongoing ICCF WC Finals there are 100% draws and thus 114 perfect games with optimal play from both sides and zero error."

see! theres another claim of hard fact by tygxc despite it being based on nothing but personal feelings and false assumptions!

tygxc you never seem to address the fact that the russians only had inferior hardware during the timespan of the interview where you cite the claim that the russians had inferior hardware from.

in fact, the literal world champion of ICCF himself pointed out how the russians did significantly worse while they had worse hardware.

"It should be noted that the lack of decent equipment is being felt quite notably in Russia, where embargoes have limited access to new hardware and the latest chess tools. As a consequence, the Russian team finished near the bottom of the standings in the last Correspondence Chess Olympiad." (2022) https://new.uschess.org/edwards-32nd-ICCF-ch

you cant help but lie in order to justify your delusions, huh.

it's also wild that you cant comprehend that people who can afford massive chess servers can just... leave the country and get the servers elsewhere, as many russians have now done.

It has no relevance, whether exactly all errors come in pairs. Of course there's always a chance for a single error game, provided that the first error is punished with flawless play. Nobody can argue against that.

What is relevant:

If paired error games are more common than single error games (would make sense according to evidence and logic), in 114 ICCF draws the expected amount of games with paired errors is >0. (Edit. probably more accurate to say it can be any positive number that is less than the total number of games in the sample)

This possibility should be accepted.

@14891

"If paired error games are more common than single error games (would make sense according to evidence and logic), in 114 ICCF draws the expected amount of games with paired errors is >0."
In the 2013 ICCF WC Finals there were 20 decisive games,
and thus 2 draws with a pair of errors are expected.
In the 2016 ICCF WC Finals there were 10 decisive games,
and thus 1 draw with a pair of errors is expected.
Now in the ongoing ICCF WC Finals there is no decisive game,
and thus no game with a pair of errors is expected.

We cannot exclude the small probability that a few (2-3) games do have a pair of errors,
but we can exclude that a substantial number (20-30) of draws contain a pair of errors,
as then at least 1 game would be decisive.

As there are 5 lines of defense that draw, the few games with a pair of errors would not change anything, as 4 lines of defense remain to draw. Thus the 114 ICCF WC Finals draws are at least part of a weak solution of Chess: redundant, but not yet complete.

"This possibility should be accepted"
++ I accept the possibility, but only with probability (1/115)² = 0.008%
If you arrive at another probability, then I would like to read your calculation.

Doesn't this rely on the idea that better engines are going to be developed which will discover errors in games played at 5 days per move with GMs supervising the strong engines that are already used?

At the normal level, most errors are picked up by the weak engine they have here in analyis. I obviously mean "most" and not "all" and I could post a game where the analysis here and "game review" dropped a win whereas playing at 3 days per move, obviously with no help from engines, I played a series of moves in a difficult ending which DIDN'T drop the win, after my opponent blundered by trying to push kingside attack for one move too many and was one move late getting his pieces back to defend. The engine here hadn't a clue how to play the resultant ending and it would only have drawn.

But all told, the chances a better engine will be able outplay a current engine enough to win are naturally falling. As we go forward in time, an engine from year y will be able to get an increasing proportion of draws against an engine from year y + 5, until radically different algorithms are developed, perhaps along the lines I've suggested or faster computers using different hardware are developed. Will quantum computers be viable enough to allow real AI instead of the "apparent AI" we have at the moment?

@14894

"better engines are going to be developed which will discover errors in games played at 5 days per move with GMs supervising the strong engines that are already used?"
++ At 5 days/move they cannot drop below 0 error/game.

"The engine here hadn't a clue how to play the resultant ending"
++ That is why the humans are still essential: they complement the engines in long term planning and strategy.

I agree.

P.S. Dio's refutation methodology. Dio thinks it goes something like

"p is false because you proposed q, which was something like p and which I refuted by telling you that you are deluded".

That's another interesting conversation.

For now we are just breaking down what can or rather cannot be concluded, in terms of total errors being made, from the fact that the sample contains 0 decisive games.

What happens when future engines, perfect or not, are introduced and meet todays ICCF finalists. Options are:

1. Future engines will be able to discover errors from the sample of drawn games we are looking at today and are decisively stronger. Possible and can't be excluded.

2. Future engines will not discover any errors in the referred 114 game sample. However, in a large sample of games future engines would beat todays ICCF finalists. They would consistently find the most challenging lines leaving less drawing lines available every turn, until our current finalists run out of depth. Eventually this would lead to an error and a decisive game.

3. ICCF finalists with 5 days of avg time for turn are strong enough to always draw against future engines. This is what Tygxc thinks. This is unlikely if not impossible depending on sample size. Reason is, we know chess games are much "deeper" than todays players are able to process. This factually leaves room for error and I see no reason to believe eventually a position would be reached where slightest of misevaluation is made, when options are as few as possible.

It's an interesting discussion, what is most likely the case today. These "options" can be broken down into more detailed possibilities.

You don't seem to understand engines. or programming. You are misinterpreting what happened here, as you often do, but usually when trying to discern the motivations of human beings.

I would remind you what Wind said only yeseterday. You seem to have forgotten.