@106109
"I put stockfish against stockfish"
++ The ICCF (grand)masters in their World Championship Finals use several different engines and with different tunings and at average 5 days per move. They intervene themselves to adjust width and depth of search, decide what moves to use more or less time on, and decide when to take their allowed 45 days of leave per year.
All of which probably makes zilch difference. I believe one previous winner said his method was just to follow Stockfish.
It's more than likely that none of the players could do anything but lose to SF, so SF probably wouldn't be too excited about their help.
If they were confident they'd include a vanilla latest version SF in the field.
@10596
"The 3-fold repetition rule is a convenient addition to the rules rather than being of fundamental interest. It settles games when neither player can work out how to win - even giving them a couple of shots to realize a position is not getting them there. Chess with a 50 move rule suffices entirely to avoid infinite play, and Syzygy tablebases suffice to deal with this."
++ It is the other way around. Chess with the 3-fold repetition rule and no 50-moves rule avoids infinite play as well: as the the number of legal positions is finite every game has to end.
The 50-moves rule is a convenient addition to ensure that over the board competitions do not go on for to long and jeopardise the tournament schedule. In the ICCF World Championship Finals an average game lasts 38 moves, so the 50 moves rule cannot trigger. The ICCF rules allow 7-men endgame table base win claims that exceed the 50-moves rule, but such claims, or even any 7-men endgame table base win claims do not occur at all. Only 7-men endgame table base draw claims happen.
On the contrary, the 3-fold repetition rule is a major drawing mechanism.
It occured in 30 of the 106 games.
@10613
"If they were confident they'd include a vanilla SF in the field."
++ The 17 ICCF (grand)masters qualified for the World Championship Finals, the vanilla SF failed to. If you think so, then enter a qualifier with a vanilla SF and see how you get beaten.
@10596
"The 3-fold repetition rule is a convenient addition to the rules rather than being of fundamental interest. It settles games when neither player can work out how to win - even giving them a couple of shots to realize a position is not getting them there. Chess with a 50 move rule suffices entirely to avoid infinite play, and Syzygy tablebases suffice to deal with this."
++ It is the other way around. Chess with the 3-fold repetition rule and no 50-moves rule avoids infinite play as well: as the the number of legal positions is finite every game has to end.
Finite but more than 50, at least before you've applied your reductions.
The 50-moves rule is a convenient addition to ensure that over the board competitions do not go on for to long and jeopardise the tournament schedule. In the ICCF World Championship Finals an average game lasts 38 moves, so the 50 moves rule cannot trigger. The ICCF rules allow 7-men endgame table base win claims that exceed the 50-moves rule, but such claims, or even any 7-men endgame table base win claims do not occur at all. Only 7-men endgame table base draw claims happen.
On the contrary, the 3-fold repetition rule is a major drawing mechanism.
It occured in 30 of the 106 games.
That's because they can't think of anything to do and they can't be a--sed to play on for another 50 moves.
And in any case the ICCF games have b---er all to do with solving chess. Not even FIDE rules and what are agreed draws meant to prove? But mainly the only indication of how the play compares with perfect play is how SF performs when you can measure it and that would suggest not remotely close.
@10591
"there are moves that are simply ignored after a zero ply analysis"
++ Yes and for good reason: game knowledge.
It is unnecessary to consider 1 e4 e5 2 Ba6? and put up a game tree for it. We know it loses.
It is unnecessary to consider 1 a4. We know it cannot be better that 1 e4, so if black can draw against 1 e4 it is trivial to draw against 1 a4.
"Trillions in a set of 102 ICCF games"
++ I estimated before that they considered 10^17 positions.
"Guessing they don't matter is acceptable (and necessary) for a chess player playing a game"
++ A human player in over the board play at 3 minutes/move average has to make some judgement calls. An ICCF (grand)master in a World Championship Finals and with engines and at average 5 days per move can go to the root.
"what you describe is trillions of nodes away from being well-described as such"
++ I estimated they looked at 10^7 positions/second/engine * 16 engines/master * 17 masters * 3600 seconds/hour * 24 hours/day * 365.25 days/year * 2 years = 10^17 positions and I estimated Sqrt (10^37 * 10 / 10,000) = 10^17 positions are necessary to weakly solve Chess.
So the effort and the requirement are commensurate.
"You are saying it needs LESS computing power to weakly solve chess than checkers!"
++ Please do not put words into my mouth. I say 10^17 positions suffice to weakly solve Chess. Most of the computer power Schaeffer used was for generating his endgame table base.
@10616
"what are agreed draws meant to prove?" ++ When both ICCF (grand)masters and their engines have lost all hope to win, then they agree on a draw. It is pointless to play on in a known opposite colored bishop endgame with more than 7 men.
"the only indication of how the play compares with perfect play is how SF performs when you can measure it" ++ Human ICCF (grand)masters + engines are much stronger than the strongest engine, and the strongest engine is much stronger than the strongest human player.
The 17 finalists qualified by winning in qualifiers.
An engine at 5 days/move is much stronger than an engine at 5 hours or minutes/move.
@10616
"what are agreed draws meant to prove?" ++ When both ICCF (grand)masters and their engines have lost all hope to win, then they agree on a draw. It is pointless to play on in a known opposite colored bishop endgame with more than 7 men.
First of all define what a "known" opposite colored bishop endgame is, then prove it's drawn, then count how many of the agreed draws are known opposite colored bishop endgames. Then at least you'll have something to say even if it is totally irrelevant to solving chess.
"the only indication of how the play compares with perfect play is how SF performs when you can measure it" ++ Human ICCF (grand)masters + engines are much stronger than the strongest engine, and the strongest engine is much stronger than the strongest human player.
The 17 finalists qualified by winning in qualifiers.
So exactly as I said - no indication there, just "I think it's pretty strong". And as I said, "Human ICCF (grand)masters + engines are much stronger than the strongest engine" would defy common sense when Human ICCF (grand)masters would lose 100% to the engine (let alone the strongest engine).
@10622
"define what a "known" opposite colored bishop endgame is"
++ There are centuries of endgame knowledge.
"totally irrelevant to solving chess" ++ It is relevant to save unnecessary calculations.
"no indication there, just I think it's pretty strong" ++ They are now so strong, that they play perfect games with zero errors. In previous years there were decisive games, each year fewer, now none. Perfection has been reached.
"Human ICCF (grand)masters would lose 100% to the engine" ++ Yes, but the engine alone would lose to ICCF (grand)master + engine, as the qualifier results prove. An unjockeyed engine is much weaker than ICCF (grand)master + engine. The ICCF (grand)master sets width and depth of search, decides how to spend his 50 days per 10 moves, decides to offer linear conditionals, sets the tuning parameters of the engines, decides when to invoke a second or third engine, when to offer or accept a draw when there is no hope, when to take his 45 days leave etc.
@10622
"define what a "known" opposite colored bishop endgame is"
++ There are centuries of endgame knowledge.
That's a pretty good definition of a known opposite colored bishop endgame.
"totally irrelevant to solving chess" ++ It is relevant to save unnecessary calculations.
If you say so, but not to solving chess.
"no indication there, just I think it's pretty strong" ++ They are now so strong, that they play perfect games with zero errors. In previous years there were decisive games, each year fewer, now none. Perfection has been reached.
No they're just behaving like SF in the examples I've given you. A few decades of engine opening analysis has severely depleted the opening phase of mates that are within its depth and the result is it draws, just as it draws in the positions you can measure where it is out of its depth (whatever the evaluation of the position from which it starts). Difficult to believe that the SF blunder rates would decrease as the number of men increases and especially that they would suddenly drop to zero when there's no tablebase.
"Human ICCF (grand)masters would lose 100% to the engine" ++ Yes, but the engine alone would lose to ICCF (grand)master + engine, as the qualifier results prove.
Then tell us how many vanilla SFs there were and how many candidates so we can judge if it proves anything of the sort.
An unjockeyed engine is much weaker than ICCF (grand)master + engine. The ICCF (grand)master sets width and depth of search, decides how to spend his 50 days per 10 moves, decides to offer linear conditionals, sets the tuning parameters of the engines, decides when to invoke a second or third engine, when to offer or accept a draw when there is no hope, when to take his 45 days leave etc.
Have you any proof it makes any difference? Sounds like you got it off the big red telephone again.
@10622
"define what a "known" opposite colored bishop endgame is"
++ There are centuries of endgame knowledge.
"totally irrelevant to solving chess" ++ It is relevant to save unnecessary calculations.
"no indication there, just I think it's pretty strong" ++ They are now so strong, that they play perfect games with zero errors. In previous years there were decisive games, each year fewer, now none. Perfection has been reached.
"Human ICCF (grand)masters would lose 100% to the engine" ++ Yes, but the engine alone would lose to ICCF (grand)master + engine, as the qualifier results prove. An unjockeyed engine is much weaker than ICCF (grand)master + engine. The ICCF (grand)master sets width and depth of search, decides how to spend his 50 days per 10 moves, decides to offer linear conditionals, sets the tuning parameters of the engines, decides when to invoke a second or third engine, when to offer or accept a draw when there is no hope, when to take his 45 days leave etc.
"They are now so strong, that they play perfect games with zero errors. In previous years there were decisive games, each year fewer, now none. Perfection has been reached."
We can't conclude that perfection has been reached. We can say that a state has been reached, where play is more accurate by both sides and the potentially existing winning lines in the games are fewer and fewer. The imperfectly calculating engines are less likely to find the rare winning option. There are far more drawing options than winning, so draw is a very likely outcome. This does not prove that missed winning lines didn't exist in ie. 100 games played - It could be than in an accurate game played by todays engines 0 winning lines occurred - or it could be 1, 5 whatever.
Now i'm stating the obvious here but this claim has been made many times here as a given.
" at 10^7 positions/second/engine "
why do you continue to make this error when it has been repeatedly pointed out?
you falsely construe nodes as a full positional evaluation.
@10616
"what are agreed draws meant to prove?" ++ When both ICCF (grand)masters and their engines have lost all hope to win, then they agree on a draw. It is pointless to play on in a known opposite colored bishop endgame with more than 7 men..
Could you tell me exactly which of the quadrillion opposite coloured bishop endings with more than 7 men are known? You do understand that in chess, the exact position of all of the pieces matters?
I created a related topic a couple of years back. Is this one "known"?
https://www.chess.com/forum/view/endgames/opposite-coloured-bishops-always-draw
" at 10^7 positions/second/engine "
why do you continue to make this error when it has been repeatedly pointed out?
you falsely construe nodes as a full positional evaluation.
I know you don't like answering me because you aren't very bright, like the others here.
But why do you continue to challenge tygxc when he has amply demonstrated that he isn't going to change what he's doing? All you're doing is making a kind of noise that the other no-hopers here will like. A bit like someone drowning and you standing on the bank shouting "save him!"
I dont answer you because you dont take the discussion seriously while pretending you do. there is no value in considering somebody who doesn't mean what they say.
" at 10^7 positions/second/engine "
why do you continue to make this error when it has been repeatedly pointed out?
you falsely construe nodes as a full positional evaluation.
I know you don't like answering me because you aren't very bright, like the others here.
But why do you continue to challenge tygxc when he has amply demonstrated that he isn't going to change what he's doing? All you're doing is making a kind of noise that the other no-hopers here will like. A bit like someone drowning and you standing on the bank shouting "save him!"
Because he's mad at tyxgc for manipulating stats
He really loves stats and it upsets him that tyxgc is scrutinizing it
At last that's what he said earlier he can correct me tho
(FIDE basic rules chess is chess without any repetition or move limit rules these days; I'm assuming you meant to say "FIDE competition rules chess".)
That is correct.