@13465
"translate any gain or loss in chess in terms of tempi"
++ That is a bold claim. Right now engines translate any gain or loss in terms of pawn units,
so with 3 tempi = 1 pawn you could translate in terms of tempi as well.
However, 1 d4 d5 2 c4 is a good move while it challenges the center, but does not develop any piece into play, so is loss of tempo for the benefit of influence on the center.
hmm
The current top engines are unable to play accurately in 6 piece endings (according to MARattigan, I think?)
...
3 piece endings.
Accurately is harder than perfectly. 5 piece endings if you mean perfectly.
They'll generally fail a randomly chosen White win in KNNvKP for example. (The average mate depth of White wins in basic rules in that endgame is around 58 full moves - unknown under competition rules, could be higher or lower.)
Last version of Rybka I got (rather old now) couldn't do KBNvK mates (4 piece).
But note that there is a lot of provably correct theory for endgames with a small number of men (including KNNvKP and KBNvK) and SF at least has tailored tweaks to it's static evaluations based on such theory. It might be expected to do better in those endgames than say 20 piece endgames where such tweaks are not possible.
hmm
@13480
"They'll generally fail a randomly chosen White win in KNNvKP for example."
hmm
@13460
"no line where that +1 tempo leads to, for example, a forced win of a pawn"
++ If there were such a line, then it has to be fast because of the dilution effect.
If it were fast, then it would have been found long ago.
"What position do you refer to when you say 1 pawn? In the starting position?"
++ The starting position or close to it, like gambit openings.
"we don't know how many pawns up is enough to win" ++ We do know.
+1 pawn near the initial position is enough to win. We know that from incorrect gambits.
"we don't know where it leads to with optimal play"
++ We know. It evaporates from +0.33 to 0.00 by the dilution effect.
"The dilution effect" is an imperfect engine evaluation in a position reached by imperfect engine play. That's the fundemental problem here that leads us nowhere in this discussion.
@13480
"They'll generally fail a randomly chosen White win in KNNvKP for example."
tygxc assumes that chess is a draw as part of the solution - so by definition whatever tygxc comes up with cannot be a solution, as its own creation assumes what it is trying to prove.
in addition, a weak solution by definition deals with inoptimal moves. not only as part of the solution tree/algorithm itself, but in addition as the moves needed to calculate what the tree/algorithm is.
"The dilution effect" is an imperfect engine evaluation in a position reached by imperfect engine play. That's the fundemental problem here that leads us nowhere in this discussion.
you'll notice how tygxc will never address this.
its also funny how optimissed cant understand the concept of a rhetorical question.
@13480
"They'll generally fail a randomly chosen White win in KNNvKP for example."
1. You confuse weakly solving with a weak solution as well as relying on your big red telephone to tell you the result of the starting position. (And in 5 of the example games I posted there were multiple blunders so after the first blunder the positions were drawn.)
2. Blunder rates in the endgame generally increase with increasing think time. A think time of 37 minutes per move as in the example I posted is hardly blitz. You're welcome to try these at 5 days per move and tell us in a few years, but note the blunder rate at 37 minutes per move was already more than 5 times the blunder rate at 1 second per move.
3. Agreed - but only for small numbers of men. Engines and humans get weaker with increased numbers of men, but humans get weaker much faster.
Fact remains - latest version of SF can't play 5 man chess perfectly.
so tygxc, wheres the deductive proof from the axioms of chess rules that chess is a draw?
the tempo valuation is a made up estimated metric. deductive proofs have no such metrics.
Let me explain it to you in short words.
No deductive proofs of chess can be made in areas where chess has not been deductively solved.
This is nonsense expressed in short words. Proofs within chess will be made in areas where it has not been deductively solved. For example a bigger tablebase consists of an unimaginable number of proofs.
Long term assessments from any complex positions, including the starting position, haven't been solved. That's what this feeble discussion is about.
True.
Therefore deductive proofs you ask for are impossible and if desired, others more inferential conclusions may be attempted.
This is an invalid reasoning: you have not provided any reason why deductive proofs are impossible, just stated the conclusion. Besides, we know that solving chess is mathematically trival - the algorithm is easily specified but impractical to execute.
They can be checked by evidence.
What on Earth is that meant to mean. 'Evidence' is not a part of a proof. It is a part of inductive reasoning (or the motivation for looking for a proof).
Where, therefore, are the deductive proofs that such inferential conclusions are incorrect? The evidence which may be used to deductively form such proofs doesn't exist?
Looks like you've tripped yourself up, are hoist by your own petard etc. Elroch just claimed that tygxc doesn't "think right for this discussion". Well it would seem that Elroch doesn't think right either, since otherwise he would be fair-minded and point out that you don't think right. You are extremely pretentious and so is Elroch, to imagine that both your ideas should be prioritised when neither of you are very bright or very right.
Both of us are bright enough not to make the mistake of using the phrase "very right". Abstract propositions are either right or wrong. Abstract truth is absolute and boolean.
@13480
"They'll generally fail a randomly chosen White win in KNNvKP for example."
tygxc assumes that chess is a draw as part of the solution - so by definition whatever tygxc comes up with cannot be a solution, as its own creation assumes what it is trying to prove.
in addition, a weak solution by definition deals with inoptimal moves. not only as part of the solution tree/algorithm itself, but in addition as the moves needed to calculate what the tree/algorithm is.
It's worth emphasizing that the reason it needs to deal with moves (by the opponent of a strategy) that are not optimal is that those moves are not already proven to be suboptimal. They cannot be ignored on the basis of an evaluation which is not only not proven reliable, but proven unreliable.
@13488
"a pawn doesn't win a game by itself" ++ It does: queen it to win. 1 e4 b5? is a white win.
"That is, in a given poisition it may be possible to checkmate your opponent if you're given four moves with no reply allowed between them." ++ Yes, but 3 tempi should be enough to win too, e.g. 1 e4 (1 tempo ahead) Nc6 2 d4 Nb8? 3 Bf4 (3 tempi ahead).
@13493
"They cannot be ignored on the basis of an evaluation which is not only not proven reliable, but proven unreliable."
++ Moves can be ignored based on logic alone. 1 e4 e5 2 Ba6? 1 e4 e5 2 Nf3 Nc6 3 Nd4?, 3 Nxe5? 3 Ng5? 3 Nh4? 3 Ng1 etc. etc.
@13488
"a pawn doesn't win a game by itself" ++ It does: queen it to win. 1 e4 b5? is a white win.
"That is, in a given poisition it may be possible to checkmate your opponent if you're given four moves with no reply allowed between them." ++ Yes, but 3 tempi should be enough to win too, e.g. 1 e4 (1 tempo ahead) Nc6 2 d4 Nb8? 3 Bf4 (3 tempi ahead).
tygxc where's your proof that 1 tempo isnt enough?
we are still all waiting for it.
"++ Moves can be ignored based on logic alone"
what logic? you still havent provided any. you just assert that a position is losing without any proof.
@13488
"a pawn doesn't win a game by itself" ++ It does: queen it to win. 1 e4 b5? is a white win.
Ignoring the awful example, the truth is that an extra pawn sometimes wins and sometimes doesn't. There are a vast number of examples of each.
Of course, you can always make up some ad hoc reason for the difference, starting from the conclusion (and if someone pretended the conclusion was different, you would come up with a different contradictory argument, showing their worthlessness).
@13490
"confuse weakly solving with a weak solution"
++ Weakly solving leads to a weak solution.
The point is to hop from the initial position to other drawn positions so as to reach a certain draw: 7-men endgame table base draw, prior 3-fold repetition, or certain draw as judged by both ICCF WC finalists and their engines.
White wins are pitfalls for black, black wins are pitfalls for white.
Optimal play by both sides avoids the pitfalls.
"Blunder rates in the endgame generally increase with increasing think time" ++ Nonsense
"humans get weaker much faster" ++ No, humans are better at long term planning.
That is why the humans play the openings, not their engines in ICCF WC Finals.
That is also why humans agree on draws in positions with not the slightest hope of winning, while engines would stupidly play on until a 3-fold repetition or the 50-moves rule.
Troitsky was better than present engines at KNN vs. KP.
"latest version of SF can't play 5 man chess perfectly"
++ But ICCF Finalist + twin servers 90 million positions/s during average 5 days can play 32 men perfectly. Troitsky analysed KNN vs. KP perfectly without any computer.
optimissed remember that arguing with feelings, as you are currently doing, isnt accepted in any sort of mathematical journals.
to all observers, optimissed is one of the main trolls. he only values tygxc's "opinion" because he too has been extensively called out on his lies by the normal people of the forum, and besides that, you'll notice how substanceless opti's statements are. it's to the point that we dont even feel the need to correct him as observers always almost immediately recognize optimissed for what he is.
you still havent addressed how you thought a guy was calling me stupid but he then confirmed that he was calling YOU stupid, lmfao.