@14304
"The 50-move rule makes the game shorter than 3-fold"
++ Yes, but finite anyway.
Besides, ICCF WC games end in true draws in average 40 moves, so the game ends before the 50-moves rule could trigger.
Chess will never be solved, here's why
Sort:
tygxc wrote:
@14304
"The 50-move rule makes the game shorter than 3-fold"
++ Yes, but finite anyway.
Besides, ICCF WC games end in true draws in average 40 moves, so the game ends before the 50-moves rule could trigger.
Whats true draws? Are you mentioning "true end"?
@14303
"rated very slightly higher in rapid / blitz / bullet" ++ rapid / blitz / bullet requires different skills than over the board classical play, which is the basis to award CM titles.
"many CM titles were awarded to weak players for winning national tournaments"
++ CM is FIDE 2200 classical rating
@14309
"Whats true draws? Are you mentioning "true end"?"
++ Yes true end certain draws: 7-men endgame table base draw, or prior 3-fold repetition, or known drawn endgame.
tygxc wrote:
@14303
"rated very slightly higher in rapid / blitz / bullet" ++ rapid / blitz / bullet requires different skills than over the board classical play, which is the basis to award CM titles.
"many CM titles were awarded to weak players for winning national tournaments"
++ CM is FIDE 2200 classical rating
Yes it was classical rating but that doesn't always prove strength in chess plenty of kids or adults who haven't played tournaments before (or barely) are underrated
Blitz chess is also severely underrated occasionally (not really a good chunk of titled players play it )
I good example would be gothamchess when he got into a time scramble his last tournament he outplayed most of his opponents during what seemed to be less than ten minutes in the clock except for two draws one which was someone rated 2800 blitz chess (I dunno what the other one is) blitz chess can be helpful to help cope with the time pressure when you think too much (people sometimes get dumb low on time it happens anyway even if you play blitz chess but it helps )
tygxc wrote:
@14309
"Whats true draws? Are you mentioning "true end"?"
++ Yes true end certain draws: 7-men endgame table base draw, or prior 3-fold repetition, or known drawn endgame.
That is not the "true end which ends in a draw". There are no best moves, computers will always improve and soon Stockfish will be replaced. There are different ways to end the game, but "true end" means a perfect game ends in: ??? We don't know what is the result after the "true end". Unfortunately 7 humans are not enough, barely 5% of Stockfish.
Jul 19, 2024
0
#10988
VerifiedChessYarshe wrote:
tygxc wrote:
@14309
"Whats true draws? Are you mentioning "true end"?"
++ Yes true end certain draws: 7-men endgame table base draw, or prior 3-fold repetition, or known drawn endgame.
That is not the "true end which ends in a draw". There are no best moves, computers will always improve and soon Stockfish will be replaced. There are different ways to end the game, but "true end" means a perfect game ends in: ??? We don't know what is the result after the "true end". Unfortunately 7 humans are not enough, barely 5% of Stockfish.
VCY looks correct.
'best moves by both sides' has never ever been established to have ever happened.
There are no 'perfect games' that have ever been known to have happened.
And Stockfish has even assigned winning advantage to positions a D player can quickly see are draws.
Are there positions where Stockfish has assigned a draw to positions that are obviously a win?
I believe there was at least one posted here - but I don't have any handy.
and tygxc is now likely to again spam that 'adding a human complements' the terrible imperfections of the computer chess engines.
No it does not.
But tygxc likely to continue to spam-claim that it does.
Whenever he does in the future.
Likely.
Soon.
llama_l wrote:
tygxc wrote:
The players and their engines complement each other.
This was true 10-20 years ago, not today.
I have no doubt there is a different level of skill regarding how ICCF players use their tools, but skill in chess playing is all but unnecessary.
-
tygxc wrote:
Seconds of over the board top players usually are weaker than the grandmaster they help too.
For example, Larry Christiansen was at one time a second of Korchnoi.
Dubov was one of Carlsen's seconds.
Notice these players are only about 100 rating points apart... stockfish on a supercomputer is about 2000 rating points higher than a 1700 player.
-
tygxc wrote:
"These players are rated below 2000 OTB"
2 are IM, 1 is FM, average age is 60, average FIDE peak rating is 2152.
I'm talking about the winner, who is rated 1900 and one of the other players I checked, who is 1700.
Jon Edwards, the winner, has no FIDE rating, and is rated 1900 USCF
https://ratings.fide.com/profile/2047080
Maybe there are still positions where engines could use a human opinion. For example in a situation where there are two available lines for which engines give a similar evaluation, but one leads to a completely closed position and the other choice leaves more open play on the board. Both are likely drawing but the human player can choose to keep the game going and chances alive.
Atleast in the past engines have struggled with evaluating closed positions. Humans could make strategic choices during the game to direct the engines to a more favourable direction, ie. more open play or avoiding lines that led to an easy draw before.
Id imagine that could involve alot of work walking through these long lines that the engines prefer and evaluating the resulting positions. At this level maybe it makes a difference in one out of a 150 games.
@14317
"there are still positions where engines could use a human opinion"
++ Many, even most. The opening to start with.
"Both are likely drawing but the human player can choose to keep the game going and chances alive." ++ The opposite also happens, the engine gives a huge advantage like +2,
but the human salvages the long term draw, e.g. a fortress.
"that could involve alot of work" ++ ICCF correspondence is a lot of work. 'In correspondence chess, he who goes to bed later wins.' - Fritz Baumbach, ICCF World Champion 1989.
That was at the start of the engine era, but still applies now.
In the pre-engine era it was hard to predict what the opponent would play.
Now you know what engines suggest to him and you can capitalise on that knowledge.
"At this level maybe it makes a difference in one out of a 150 games."
++ It makes a difference in many more games.
That is how the Finalists qualified through Preliminaries, Semifinals, Candidates:
by winning against players with access to about the same hardware and software.
Even more: 4 of the 17 Finalists are Russian and have inferior hardware because of sanctions.
@14315
"There are no best moves"
++ There are only drawing moves and losing moves, non-errors and errors.
"computers will always improve" ++ They cannot improve below 0 error/game.
They will reach 0 error/game in shorter time than 5 days/move.
"There are different ways to end the game" ++ So far in ICCF WC Finals there are 3 ways to end a game: 7-men endgame table base draw, 3-fold repetition, and agreement.
"a perfect game ends in: ???" ++ The 113 perfect games in the ICCF WC Finals end in draws in average 40 moves, standard deviation 11 moves,
the shortest 15 moves a 3-fold repetition, the longest 73 moves also a 3-fold repetition.
"what is the result after the true end" ++ A draw
"7 humans are not enough" ++ 17 humans each with twin servers of each 90 million positions/s and at average 5 days / move.
@14291
"the engines are rated 1000 points higher"
++ There is more difference between correspondence 5 days/move and classical 3 minutes per move, than between classical and bullet 1 s/move. It is not always the top classical rated player who wins bullet tournaments. Different time control, different game, different skills.
Boris Vainstein, not even a master, was the second of Bronstein in the Budapest 1950 Candidates' Tournament. Prior to the resumption of this adjourned game Vainstein showed the final position to the Soviet ambassador: he had analysed it all the way overnight.
https://www.chessgames.com/perl/chessgame?gid=1033779
Yury Dokhoian was many years the second of Kasparov, though he was much weaker.
Analysis and play require different skills.
@14321
"a few ICCF final losses I posed a while ago were simple blunders" ++ Yes, clerical errors happen, time trouble happens too: 50 days/10 moves, but when all is used up they must reply the same day, also computer crashes, real life problems, illness...
"some players made it to the finals without using an engine" ++ No.
"beating a non-zero number of top level ICCF players is not that hard"
++ You have to do it in Preliminaries, in Semifinals, and in the Candidates.
"immediately recognized one of the games was lost (the player resigned on the next move)."
++ I guess that was time trouble/sloppyness, but the refutating queen sacrifice is pretty.
https://www.iccf.com/game?id=1164336
Here are 2 games won in the opening due to the long term queen's side advantage:
https://www.iccf.com/game?id=948179
https://www.iccf.com/game?id=948250
Here another one won in the opening:
https://www.iccf.com/game?id=278961
There are 2 main reasons why they now reached all draws: care to avoid clerical errors,
and avoidance of dubious openings refuted in past ICCF games.
"These are not strong players"
++ But strong analysts, strong in positional play and long term planning.
"These are weird people" ++ Mostly retired, with much time and patience.
"with a strange hobby" ++ Chess is a strange hobby.
Correspondence is no more strange than classical, rapid, blitz, or bullet.
llama_l wrote:
Kotshmot wrote:
Maybe there are still positions where engines could use a human opinion. For example in a situation where there are two available lines for which engines give a similar evaluation
I think that would be a very bad way to do it. A human can only look through a limited number of lines anyway. Better would be a Monte Carlo (or similar) method. Have the engine play out an enormous number of games that go down different branches, then compare the results.
The skill would probably be related to optimizing the tradeoff between depth and number of games. For example you could have an engine play many games per second. but only at a low depth.
-
Kotshmot wrote:
Atleast in the past engines have struggled with evaluating closed positions. Humans could make strategic choices during the game to direct the engines to a more favourable direction
Yes, in the past that was true.
Based on my experience of disagreeing with the engine years ago but not these days, and based on things said in interviews with Carlsen and Caruana, I don't think humans can add anything anymore. For top players the paradigm has changed. Instead of ignoring the engine when it suggests something objectively wrong, they ignore it because in spite of being correct it's impractical because they can't understand.
Caruana told an interesting story on the c squared podcast about banging his head against a position, not understanding why it was supposedly winning. Finally he had this idea, to have a lesser engine (something like 3200 strength) play it out to see how it would win... but that engine couldn't win it (only draw). After that Caruana gave up trying to understand why it was a win because a player much better than him (3200 engine) couldn't do it either.
Sure, a choice based on just human understanding alone is not optimal - Although human chess understanding would still be involved identifying the critical positions and lines that are worth investing time in. Like in the example I gave earlier - a line that leads to a completely closed position might not be worth investigating every branch with high depth.
Jul 19, 2024
0
#10995
MARattigan wrote:
MEGACHE3SE wrote:
ngl mar ur comment may be pretty confusing/misleading unless they know the context better. ur making it seem as if we already know it, not that we know the algorithm to obtain it.
The algorithm is the solution.
The time efficient algorithm is, "look up the position on the syzygy site and play the top move". It's a lot quicker than going through the tablebase generation process, but that just makes the latter a less time efficient solution.
The definitions of solutions should include the maximum time (or complexity) for the algorithm to complete as a parameter. As it is the tablebase generation algorithms and the minimax algorithm, without time limitation, both satisfy the definitions that you can generally find of weak solutions (and could be extended to strong solutions). There is usually some vague accompanying text about time, that could mean anything, but it's not included in the definitions.
As the lawyers like to say, "time is of the essence".
I should also have mentioned space perhaps. The tablebase generation solution would run out of storage space in any practical (but actually totally impractical on the basis of time) attempt to apply it. (Though I think it could be amended to avoid that at the expense of a vast increase in the already vast time required.)
tygxc wrote:
@14315
"There are no best moves"
++ There are only drawing moves and losing moves, non-errors and errors.
"computers will always improve" ++ They cannot improve below 0 error/game.
They will reach 0 error/game in shorter time than 5 days/move.
"There are different ways to end the game" ++ So far in ICCF WC Finals there are 3 ways to end a game: 7-men endgame table base draw, 3-fold repetition, and agreement.
"a perfect game ends in: ???" ++ The 113 perfect games in the ICCF WC Finals end in draws in average 40 moves, standard deviation 11 moves,
the shortest 15 moves a 3-fold repetition, the longest 73 moves also a 3-fold repetition.
"what is the result after the true end" ++ A draw
"7 humans are not enough" ++ 17 humans each with twin servers of each 90 million positions/s and at average 5 days / move.
True or perfect endings are not possible. You can't tell the difference between e4 and d4.
The true end will never be achieved. There are more than 2 best moves that we will never tell the difference, thus the true end is not possible. You don't understand what "true end" we are talking about.
@14326
"True or perfect endings are not possible" ++ They are.
"You can't tell the difference between e4 and d4" ++ No difference, both draw.
"The true end will never be achieved" ++ It is being done now: 113 draws out of 113 games.
"There are more than 2 best moves" ++ 1 e4 and 1 d4 are played now, also 1 Nf3.
1 c4 was played in the previous ICCF WC Finals.
"You don't understand what "true end" we are talking about." ++ Please explain.
tygxc wrote:
@14326
"True or perfect endings are not possible" ++ They are.
"You can't tell the difference between e4 and d4" ++ No difference, both draw.
"The true end will never be achieved" ++ It is being done now: 113 draws out of 113 games.
"There are more than 2 best moves" ++ 1 e4 and 1 d4 are played now, also 1 Nf3.
1 c4 was played in the previous ICCF WC Finals.
"You don't understand what "true end" we are talking about." ++ Please explain.
Explain how a "true end" may be achieved. ICCF WC is not a valuable source, it isn't apart of this party.
Jul 19, 2024
0
#10999
I wonder if the supercomputers have tried playing each other where white opens 1) g4 and if so - what happens?
g4 is arguably one of the worst first moves or perhaps Thee Worst.
Found this page on Reddit - but nothing about 1) g4 yet.
https://www.reddit.com/r/askscience/comments/3jmb3f/what_would_the_result_be_if_2_perfect_chess/?rdt=53432And tygxc is of course wrong again ... this time about 1) e4 versus 1) d4.
He doesn't know the outcomes but tries to claim there's no difference because he falsely claims he knows what the outcomes are.
But of course he's been spamming such circular invalid arguments the whole two years.
Spamming but not necessarily trolling because there is some difference although spamming usually comes under the heading of trolling.
He's probably going to spam-claim shortly that he 'knows the outcome' of 1) g4 ...
Likely that he will. Soon.
Jul 19, 2024
0
#11000
tygxc wrote:
@14291
"the engines are rated 1000 points higher"
++ There is more difference between correspondence 5 days/move and classical 3 minutes per move, than between classical and bullet 1 s/move.
This is questionable but actually underlines the point. Although the ratio of times is higher - 2400 times more time versus 180 times more time - there is a falling rate of increase in rating per doubling of compute time. But a modern engine with 7200 minutes per move is more than 1000 points stronger than any human with the same time. I guarantee it would beat any human 100% of the time in such games.
It is not always the top classical rated player who wins bullet tournaments. Different time control, different game, different skills.
So what? The right rating scale is the one for the time control.
It is interesting that Blitz and Bullet chess have a significantly wider rating scale.
Note that the ICCF rating scale has suffered massive, invisible inflation since engines became of comparable strength to humans. The reason is simple. Pretty much everyone kept getting better engines at the same time, so the changes scarcely affected their ratings at all. They are like engine ratings where the engines got replaced by inherited the rating of the previous version. In fact, the top of the scale has even been compressed a bit - there are a few retired and deceased players who have ratings higher than anyone now. These date from the era when the best humans could contribute something. Ulf Andersson was 2820 in 1995, but his rating had fallen by the time engine ratings were exceeding those of all humans in 2004. Here is how CCRL ratings went from a couple of years later (while ICCF ones did not rise).
Boris Vainstein, not even a master, was the second of Bronstein in the Budapest 1950 Candidates' Tournament. Prior to the resumption of this adjourned game Vainstein showed the final position to the Soviet ambassador: he had analysed it all the way overnight.
So?https://www.chessgames.com/perl/chessgame?gid=1033779
Yury Dokhoian was many years the second of Kasparov, though he was much weaker.
So?
Analysis and play require different skills.
Stockfish is very, very good at both. And imperfect in ways humans can't help with.
@14291
"They download the games and have some program run a batch process on it to see if their setup disagrees with any move, and if so, they try to repeat that opening?"
++ Do you have the same low opinion of the seconds of over the board grandmasters who prepare the openings of their grandmaster?
I enjoy both chess and data.
For example recently I had some fun re-creating this graph
https://www.reddit.com/r/chess/comments/v3mscu/i_analyzed_chesscom_blitz_rating_distribution_for/
That one is 2 years old, and here's the same one today showing how the averages have changed.
-
However I found out our methedologies were different. The first grpah used the leaderbord, while I used all titled players on the site.
The other user wanted to avoid inactive ratings. My method for avoiding inactive ratings was to make another graph (not shown) which filtered by RD (which made the ratings quite a bit higher than either graph).
Also I think median or mode is more interesting than average since these are not gaussian...
... anyway, my point is I can respect both types of skill. Skill in chess and skill in data processing and interpretation... but I do not confuse them as being the same thing.
where are the y axis labels
y-axis is in the left edge of the label, it doesn't have a value.