Well you still haven't shown us your win as Black against SF after 1.e4 e5 2.Ba6. Why don't you do that first before continuing to post that you can tell what lines are random and losing.
It would give you some credibility.
If you could do it against something that can mate with a rook and king against a lone king that would be even better, but let's see how you do against SF first.
@5451
"Those who contribute to the peer-reviewed literature say chess is too complex to solve."
++ I adhere to the peer-reviewed literature and say humans and computers can weakly solve in 5 years. Peer-reviewed literature accepts the use of knowledge. Peer-reviewed literature accepts brute force as well as rule based strategies, so accepts acombination of both as well.
Clear win / loss / draw is adjudicated by the humans based on rules like Allis' solution of Connect Four. Unclear positions are calculated by brute force until a table base, or a prior 3-fold repetition, or a clear draw / win / loss that humans can adjudicate based on rules.
You say "no, let's change the meaning of solve, and find ways to make it harder to solve and then we can say we cannot solve it". I say it can be solved with 10^17 relevant positions. You say "no, let's find arguments to make that figure higher" I use thought to reduce the number to 10^17. You use thought to increase the number.
"A concrete example is 1. e4 e5 2. Ba6. Regarding the weak solution of chess (same meaning as the entire peer-reviewed literature), this is an unproven case"
...
Solve chess without the 50-moves rule.
Then that same solution also applies with the 50-moves rule.
Only if you're a moron.
But the topic our foregoing exchanges was whether the tablebases strongly solve positions with 7 men or less. You seem to be trying to change it.
I tend to think that tygxc is right on that one. At some point too, it's necessary to relax the emphasis on deductive reasoning. As I pointed out to Elroch, you can't solve chess without the mind of a scientist and for all the riduculous stuff about five years, tygxc still thinks more like a scientist than most of the others. A scientist with an unfortunate obsession, maybe, like Benny Hill in The Italian Job. Of course, I doubt Elroch believes me but nevertheless, with a mindset in which we can't know if 1. e4 e5 2. Ba6 loses for white, science isn't being allowed to play a part. It's as though he and others have never heard of successive approximations and closing in on an accurate result incrementally. No, for them it's got to be all worked out deductively. That's rubbish.
Let's be clear here. This is not an objective disagreement (the interesting kind). It is a worthless semantic disagreement. "Worthless" because if you avoid using the same word for two different things, there is no disagreement. Those who contribute to the peer-reviewed literature say chess is too complex to solve. You and @tygxc say "no, let's change the meaning of solve, and then we can say we can solve it". Using the same word for two different things and then failing to acknowledge you are doing this is obfuscation, and its only contribution to objective knowledge is to make it more difficult to communicate about it.
A concrete example is 1. e4 e5 2. Ba6. Regarding the weak solution of chess (same meaning as the entire peer-reviewed literature), this is an unproven case requiring proof if this position can be reached from either of the two candidate strategies.
All of us agree it is very likely a win for white. Some don't understand that it be an excellent bet - maybe one you could stake your life on - but not epistemiologically justifying certainty. They erroneously think that induction from the tiny amount of existing chess praxis by flawed humans and engine including, say, an evaluation of 500 centipawns (like many positions that are not won) and a LeelaZero evaluation of 99.8% (or whatever it is) is enough to be certain. No, it ain't. Maybe you could stake your like on it, but staking your life on 10^20 such examples all being correct would be suicidal.
This is Elroch's effort. The same thing again about 10^20 "such examples". We're not talking about 10^20 "such examples", Elroch ... we're talking about that specific example, so who is currently living in fantasy land? Do you think that 1. d4 might lose by force? You'd really be forced to think that, if you think 1. e4 e5 2. Ba6 might not lose.
Few people have a good intuitive understanding of the quantification of belief: all studies show systematic errors in how people deal with extreme cases. Perhaps I can help someone (but only those conducive to improving understanding)
An excellent way to think of the meaning of probabilities is how they relate to returns on bets - the way it was first thought of in the 17th century. They asked the question "would it be wise to take a bet at certain odds?"
For the question of whether 1. e4 e5 2. Ba6, an assertion of certainty that this is a black win is a belief that the correct odds for this to be a win for black are 1:0. This would mean, for example, that it would be an excellent bet to stake say, the entire wealth of the planet versus one unit of some absurdly inflated currency - say worth a billionth of a cent - on it being a win. Or a bet far more extreme than that, with some more imagination!
True certainty would make such a bet a no-brainer - there is literally no risk and there is a finite reward, so it should be taken. But what you see as "near enough" certainty makes a sufficiently extreme bet foolish. That is the real situation.
@5174
"But what you see as "near enough" certainty makes a sufficiently extreme bet foolish."
++ Betting is about probability of an event.
You could bet about me winning 1 e4 e5 2 Ba6? as black against Carlsen.
Solving a game is not about probability, it is about certainty.
It is deterministic, not probabilistic.
It is certain that 1 e4 e5 2 Ba6? loses for black with best play from both sides.
It is certain that 2 Ba6? does not oppose to the draw more than 2 Nf3.
It is certain that 2 Ba6? is not optimal play.
I have proven @5173 that it is checkmate in 82 but that is not even necessary.
Black is up material. All other factors are equal. Hence white loses.
It is certain that 1 a4 opposes no more to a draw than 1 e4, or d4.
It is certain that 1 Nh3 opposes no more to a draw than 1 Nf3.
It is certain that 1 a4, or 1 Nh3 cannot be more optimal than 1 e4, 1 d4, or 1 Nf3.
It is certain that the endgame https://www.iccf.com/game?id=1164259 is a draw.
In 1976 some mathematicians argued that a proof could only contain human deductive arguments and they rejected the computer proof of the
https://en.wikipedia.org/wiki/Four_color_theorem.
It seems some people here only accept computer output as proof
and not human deductive arguments.
@5154
"an evaluation of 500 centipawns (like many positions that are not won) and a LeelaZero evaluation of 99.8% (or whatever it is) is enough to be certain"
++ No, engine evaluations are flawed.
Certainty comes from calculation until the 7-men table base or a prior 3-fold repetition.
However, certainty also comes from human deductive logic.
When all other factors are equal any material advantage wins.
After 1 e4 e5 2 Ba6 black is up material and all other factors are equal, so white loses.
The correct method, if an attempt is to be made to solve chess, is first to try to develop a theoretical understanding of points of flux in chess games. That is, points of tension which are complex and where small variations can mean the difference between winning and losing. I would call them positional-tactical melees, perhaps, where tactics can change the positional sense of a game. They need to focus entirely on that aspect of chess for several years and try to build a store of pattern tranformation recognition. Are you with me? Then they need to try to perfect algorithms dealing with just those game aspects and no others. Gradually, over time, it will be seen that the understanding of the influence of tactical exchanges becomes more wide-reaching. I think that solving chess is hampered by the existing focus on entire games. Maybe I should contact @Caproni, to see what he thinks of my idea. If he says that there's a problem with my thinking, then I'll believe him, because I would know he would have thought it through.
Here we are again at the crux of the disagreement. This method of solving chess relies on using the judgement of GMs or engines that are admittedly imperfect to determine which elements of the game are important and which are irrelevant. Even getting consensus among GMs on which opening lines are best is impossible--witness the many times we see players choose a line that their opponent has declared to be inferior, just to prove them wrong. And of course any engines used would have been surpassed by new developments before the study could be completed, casting doubt on the entire process.
I continue to believe that only a brute-force computation of all possibilities can be entirely reliable. I also believe that such a task is impractical at present, but likely doable at some future time.
dang u right
I, for one, understand your preference but would respectfully suggest that a brute force analysis of the entirety of chess is impossible for the forseeable future. Therefore it seems that, after a brief flirtation with the idea of perfection as embodied in more and more powerful computers, we're once more thrown back on our own resources. Therefore it is perfectly acceptable and reasonable to claim that a lost position, such as 1. e4 e5 2. Ba6, is well and truly lost.
The first thing you need to accomplish a task is the tools and materials needed for it. As we both agree that the foolproof solution is presently unavailable, we can speculate and give our best opinions, but best guesses so often prove to be incorrect in the long run that that any quick solution at which we might arrive must remain doubtful.
@5154
"an evaluation of 500 centipawns (like many positions that are not won) and a LeelaZero evaluation of 99.8% (or whatever it is) is enough to be certain"
++ No, engine evaluations are flawed.
Correct. They are sometimes misleading (although of course the AI ones admit their uncertainty, so there is no excuse for thinking otherwise.
Certainty comes from calculation until the 7-men table base or a prior 3-fold repetition.
However, certainty also comes from human deductive logic.
If you understood what deductive logic was, it would help. It can be expressed as the repeated application of a very small number of syllogisms. This could not be mistaken for your glib non sequiturs.
When all other factors are equal any material advantage wins.
After 1 e4 e5 2 Ba6 black is up material and all other factors are equal, so white loses.
LeelaZero's billion parameters for evaluating everything about a position (trivially including the material and anything you might include in "all other factors" (plus a million times more) provides it with enormously more testable understanding about this but does not provide it with certainty. A passable human player like yourself being certain about this is an example of your poorer judgement versus an AI that is over 1000 points stronger.
...
It is certain that 1 e4 e5 2 Ba6? loses for black with best play from both sides. (really?)
It is certain that 2 Ba6? does not oppose to the draw more than 2 Nf3.
Ergo 1 e4 e5 2 Nf3 is a win (for someone).
...
It seems some people here only accept computer output as proof
and not human deductive arguments.
I can see their point.
@Optimissed re:
That is a ludicrous misrepresentation.
Certainly what @tygxc says at any rate. In your case it should be "let's change the meaning of solve, and then we can still say we can't solve it".