Chess will never be solved, here's why

#2280

"So in your judgement he'd have played different moves depending on whether it'e 2-fold, 3-fold etc. Is that what you mean by " It does not matter"?"
++ They played 40 moves in 2.5 hours. Of course Kotov used all his available time calculating 21 moves deep before he sacrificed his queen 30...Qxh3+. So he took advantage of the 3-fold repetition rule by repeating twice to reach move 40 without losing on time. If it had been a 2-fold repetition rule, then he would not have repeated twice, or he might have lost on time, or he might not have sacrificed his queen.

When I say it does not matter I mean it does not matter for solving chess. If repeating twice are the optimal moves, then repeating 3-fold are the optimal moves as well.

#2282

"I think rook or knight would lose." ++ Yes, that is correct.

"The point is, according to your own description of your method your computation must consider the resulting position where White has two dark square bishop's. Precisely what you say will never happen."
++ I do not count positions with two dark square bishops in my assessment of the feasibility of weakly solving chess.
Of course I allow any underpromotions that may arise during the actual weakly solving of chess.
This is by the way no counterexample to the heuristic of never underpromoting to a bishop unless to avoid stalemate. Promoting to a queen is still the simplest and best way to draw.

"A heuristic (/hjʊˈrɪstɪk/; from Ancient Greek εὑρίσκω (heurískō) 'I find, discover'), or heuristic technique, is any approach to problem solving or self-discovery that employs a practical method that is not guaranteed to be optimal, perfect, or rational, but is nevertheless sufficient for reaching an immediate, short-term goal or approximation. Where finding an optimal solution is impossible or impractical, heuristic methods can be used to speed up the process of finding a satisfactory solution. Heuristics can be mental shortcuts that ease the cognitive load of making a decision."

'self-discovery'.  Lol.  'Ease the cognitive load'.
But 'decision' seems to have already been made.
This is about Sveshnikov ?    
Its about 'fighting for Sveshnikov'  ? 
Interesting that 'heuristic' can refer to 'self-discovery'.
Plus we got a few posts ago 'can't help it'.

Anyway - 'heuristics' would certainly be relevant to the forum subject..
And - to a big percentage of the postings so far.

Repeating any number of times does not matter in a weak solution of chess, because the repetitions can simply be removed.

Whether it matters in computing a weak solution of chess depends on the method you use.

If you use a tablebase generation method it doesn't matter.

If you use your proposal it does matter.

That's because SF14, your vehicle, will repeat moves (possibly optimally) in won drawn or losing positions. (Not more than twice if it evaluates alternative moves as winning). 

For example the 1024 sec/move example in the SF14 v SF14 games I posted here was the only one that finished correctly in a win from the initial winning position. It was played with a 3-fold repetition rule in force. Had it been a 2-fold repetition rule the result would have been a draw.

"If repeating twice are the optimal moves, then repeating 3-fold are the optimal moves as well."

Unfortunately typical of your reasoning. White's moves 4 in the 8 sec/move and 256 sec/move games in the same link repeat twice and are perfect in your 3-fold variant. Had he repeated 3-fold on move 6 that would have been a blunder. 

Both those numbers (10⁹ and 10⁷) were referred to the proof size, not the number of nodes searched and I think you know it well now as before. You have been saying that only 10⁹ or 10⁷ nodes had been searched to solve checkers.

And is it impossible that's not enough to make it understandable, not only to the less gifted?

You stil put many things together, as if they were questionable the same.

There is no possibility to construct anything, as long as you think that any other either cannot or doesn't want to understand you. You simply do not accept that basically all the others may have the same opinion about your theories and you attribute that to "ganging". That's a clear sign of hubris imho.

As I said, you exclude the possibility that your reasoning may be impossible to understand for anyone.

Capablanca did not graduate at university, if I am not mistaken, but yes he did study engineering. Nonetheless Tal didn't, for example; he studied psychology, if I recall well. Morphy and Alekhine were lawyers, Kasparov has a degree in literature, I think. Why do you mention only facts which support your theories, when you can easily find others that do not? And I hope you believe my word that I have been outplayed at chess by people who had far less scientific education (and understanding) than me.

Incredible. You really think you "beat" all the others here at science, which is about objectivity, when you, a single subject, entitle yourself and yourself only to evaluate how much scientific you are... And for sure I do not see this thread as a contest; you do, and you are so confident this is the only way to deal with such things, that you think the others do the same. A true narcissist. @playerafar was right, then, you behave like an unscrupulous lawyer: first the win, then the truth.

I edited my previous post to make myself clearer: you think you can use your supposedly better than ours evaluation function, to understand how much of the search space can be cut down, preventing any further search in that excluded part, in order to speed up the process. AFAIK no one out there (not only in this thread) supports your approach, because it is fallacious by faulty generalization.

I'll take that as a "yes".

Sorry for the delay, I didn't at first realise that was meant to be a response to my challenge.

Very good as a first try, but a few nit-picks.

(i) If you add moves and/or change the move order some people might say that changes something by definition.

(ii) You don't prove the accuracy is near 0.

(iii) You don't prove if you add some moves and/or change the move order, the accuracy stays near 0%.

Apart from that - well done!

I look forward to you clearing up those few points - I'm sure as the World's strongest grandmaster it won't take you long.

Apparently you don't understand how computers work.  If you give the system a set of rules to apply to "eliminate" non-sensible positions, the act of evaluating each position and checking the criteria you have set is a significant portion of the effort you think that you are avoiding.    Whether you pre-parse all this and de-couple it from the actual evaluation of the sensible positions make little difference...you're still talking about orders of magnitude that make your 5 years laughable.

10^120 possible games including all legal paths *was already reduced* to 10^44.5 unique positions since the route of evaluating all legal games is even more out of reach.  The cost of this simplification is that you are left evaluating unique positions, not valid game paths.  Checking those game paths will cost you a number of years that are not remotely feasible.  As I said, your reductions are not going to work.  10^120 to 10^44.5 (sound) to 10^36 to 10^17 (unsound)...it's a little much .

yes

The difference is, I have demonstrable credibility on the forums .  Every would-be genius on the forums is the same...they all think that their brilliance is unfathomable and leads to people "ganging up on them" in some organized way.  The answer is much simpler.  You're just wrong, ergo a bunch of people individually attempt point it out to you. 

Optimissed tried to name us all as a group, but Elroch and I don't really interact directly, Playerafar and I don't really get along, Martiggan and I have never interacted that I can recall, and Haiaku is brand new addition to this topic (and a very discerning one, I would say).  Pfren, BlueEmu, and Llama each have their own unique take on things as well.  There's no "gang".  If the average user on the forums understood this problem better, you'd have hundreds and hundreds of posters telling you you are wrong.  How big does the "gang" of posters telling you your premise doesn't work have to get before you even start to wonder if you *might* be mistaken?

The fact that it happens over and over should clue you in, but as we know from other examples, some people are not that good at picking up on clues that they might be off the beam.  

Congratulations.  I accept your premise that this position is not the solution to chess .  Only 10^44.5 more positions to go...

No, I did not start that crap .

Yeah, I seem to be less hostile in this topic. I'm willing to consider how chess may or may not be solved in the future.

The main hurdle for me is one you've brought up a lot (I haven't been reading this topic religiously, so I don't know if @tygxc has answered it yet) and that is saying out loud that a class of positions can be ignored is very different from devising an approach that allows the software to completely skip over them.

From a distance, the argument seems to be that if we let a computer catalogue a large number of positions we can slap a "weakly solved" label on it. I mean sure, ok. As far as I know there's been no strict definition given to "weakly solved." You want to say that an engine plus enormous catalogue would outperform today's engines? Sure. Sounds obvious. If you want to say more than that then let's hear it

Promotions to bishops may be the rarest, but they empirically occur in 1 in 33,000 games i.e. well over 1 in 10,000,000 moves, which is likely not too different to the frequency of positions where such a move is optimal. That is plenty to use the intuitive argument I provided why a substantial fraction (something like 0.1%) of all legal positions with multiple underpromotions are reachable via optimal play.

The argument that promotion to a bishop can only be strictly superior to queening in a winning position sounds correct.  However, I believe eliminating lines where an opponent can win by promotion to bishop could be crucial to analysis of a drawing position (in order to avoid getting into them -  assuming the opponent can only queen could be fatally deceptive). That makes such positions potentially important. The same for positions with 5 underpromotions to bishop to achieve a win.

Yeah, if the idea is that the weak-oracle can't play optimally in every legal position, but it's impossible for it to lose when playing a game from the starting position, then I don't know how you can ignore underpromotion.

If the idea is that the weak oracle will almost never lose, but it would be impossible to meaningful improve its play short of fully solving chess, then I don't know how you can ignore underpromotion.

If the idea is that the weak-oracle will almost never lose, and future improvement is possible... well ok, but that's just a strong chess engine. We're already cataloguing analysis. Analyze a well known position on lichess and the depth is instantly at 40. Load up live book on chessbase and some positions have analysis saved to depth 60+.

"The fact that it happens over and over should clue you in, but as we know from other examples, some people are not that good at picking up on clues that they might be off the beam"

He's already asserted - twice now that he's 'pretty sure he knows more than anybody here' ...
its more than 'not that good'.
Its 'serving notice'.   And it explains his reactions and policies.
You don't tell him something.  He tells you something.
Tell him anything - and we'll get another column of unfounded assertions instead of acknowledgement. 
He'll even try to claim the speed of the computers doesn't matter.
Has.  Vehemently.

Is this all to 'vindicate Sveshnikov' ?
I don't think so.
This is not about 'weak solving' either.
Its about 'heuristics'.   

Unfortunately, that was the fly in the ointment. It's possible to allow a computer to follow all the possible chess lines to their ultimate conclusion and, several hundred million years later, the computer would stop because its job would be done.

The representatives of mankind, whatever they looked like, would look at the computer, stopped: and probably think it signifies the end of the Universe. That's because there would be no decipherable output. That's because, if the computer was programmed without assumption that chess is drawn, (or that chess is a win) then it couldn't discard any lines that are the negation of the assumed outcome. It couldn't really discard any lines at all, in any case, because, to put it into simple terms, it couldn't know where one line stopped and others began. There would be no differentiation and that is the logical product of a "strong" search, whereas a "weak" solution can't be produced without extremely powerful algorithms, such as have not been designed.

This brings the entire conversation back to some point in Ponz's thread, where I proposed that a novel approach is necessary. Everything since that point has been off track. The novel approach involves identification of types of position, rather than following specific positions. It was claimed that type classifying is impossible because chess is all about the concrete and the specific. Well, if that were true, everyone is wasting their breath, because the project is impossible, given present computing speeds.

One mistake that is being made is this: in criticising tygxc's belief that Stockfish provides the algorithms and the engine follows them to some database position or other. It's been claimed that assessing each position requires just about as much computing power as reaching such a position does. Actually, it requires far more and this is what makes the project realistically impossible. It is actually better to analyse games rather than positions, because the amount of computer time needed to assess every position is equivalent to following a game anyway and therefore, a game led approach cuts down on computing time, rather than increasing it.

All this stuff about numbers is guessing, by the way. Everybody is, of course, an expert, who authoritatively produces figures.

It's why it's so funny.

I am sure that tygxc isn't the only one in error though.