Chess will never be solved, here's why

@MARattigan
Thank you for that reply.
It might be good to scroll down to these two lines and adress them first:
You also posted  "Confirmed - I posted examples for you"
The key word there is 'Confirmed'.
One might be tempted to post in-between text before seeing and considering the further context or whatever further below.

"FIDE doesn't refer to castling as a sideline - not in the laws at any rate. You'll need to explain that."
I didn't say that.  I said something else. 
I don't need to explain something I didn't say. 

Regarding your diagrams that you've posted twice - with two Kings and a rook - 
I appreciate you taking the time to do the diagrams - 
but perhaps it would be better if you simply articulate your points about that in the fewest possible words.
If you really do understand it - you would have no difficulty doing so probably.
But please read further first - before considering response.

You also posted  "Confirmed - I posted examples for you"
The word there is 'Confirmed'.
I attempted to paraphrase the point or points that you seemed to be attempting in several posts.
But then you said 'Confirmed'.
Is that the point or points you're trying to express through the diagrams?
If you Confirm that also - then we could see that as progress.
I articulated it verbally.   Is that now 'taken care of' to now be built on ?

Regarding multiplying positions by 100 - as 100 positions because of the possibility as to where they stand in the 100-ply 50 move process ...
I believe we could make further progress with that.
If you're suggesting that's necessary for each position in chess -
and it actually is - then so be it ...
but I'm suggesting that needs further evaluation as to whether to implement that or discard it or leave it in paradox and controversy.

Then maybe you should check numbers better, instead of firing them like a machine gun; deal adequately with whichever of the many objections moved to your facts and figures, before moving to the next one, instead of trying to "beat" your opponents in simul with your "succint" but insufficient explanations, don't you think?

(The words I edited in red are clearly a repetition) We too have patience, and you appear to us unable or unwilling to understand. The problem is not that number per se; it is that you will know (and when I say "know" I mean "by proof") how much of the hypothesized search space has been searched, after the search, not before. You, instead, make basically no difference between your hypotheses and facts not yet occurred.

I for sure do agree with you on that and never stated the contrary.

Well you see you are no worse than anyone else at insulting. Some of your insults are more oblique, like the allusion at people's understanding and their ratings. And as I said, your way to mix facts and opinions as they were the same thing is insulting per se imho.

I think that no number of opponents would ever convince you that you are wrong, so I wonder what criterion you use, other than yourself, to prove you are wrong. Of course, you could be right and all the others wrong, but people found (or think they have found) an awful lot of errors in your theory, not just one or two, and you think all them are not objective while you are. Fine, but I think we all have to discuss more about how "objectivity" can be defined too, otherwise anyone can say s/he is more scientific than anyone else. It would be quite a paradox, you see.

I think you actually do a couple of those things, but a part from accusing, is that what unscrupulous lawyers usually do?

Common sense is an evaluation function and heuristics are nothing special: they guess no less, but you want to use them to decide a priori which nodes are not worthy of calculation.

It's shocking that you are shocked, because I am not just one of those who thinks it must be done, in order to claim chess weakly solved: I think you are basically the only one, in and out this forum, who thinks otherwise.

After spending 60 hours to add to the search only 4 candidate moves for white and one child for each of them (so you said). So after five years how many nodes will have been searched?

     If you feel that it is inappropriate to talk about such matters, WHY do you continue to do so?

Its apparently as a baiting manoeuver - and to push for a situation whereby there is a double standard for him.  A power play.
As to whether he has an inkling of his own motives -   
The other guy continues to 'look good' by comparison.
'Soft guy Hard guy' dynamics.  

It must be really irritating to have people being pompous when they know so much less than you, who understands everything without even trying; not being able to explain why you are always correct to the chattel must be even more frustrating ...it's such a difficult burden you carry.

Hard Guy doesn't understand that it was he whose 'feelings' were affected when he was reminded that its always a mismatch with Elroch and others always coming out on top of him.  His reference to IQ was intensely feeble - compounding his persistent errors.

We can continue to be polite to @tygxc - and not indulge in the namecalling that the moderators had apparently cured the other guy of temporarily.
Please notice that the terms 'Hard guy; and 'Soft Guy' don't refer to usernames.
Nor do they have to be pejoratives.
Soft and hard can be plusses in many situations.

Apparently - @tygxc was doing FORTRAN in the seventies.
So he isn't 'young' after all.  
That could mean that he isn't going to change either.
He'll continue to hold that he 'knows more about mathematics' while also continuing to hold that the real speed of the computers 'doesn't matter' and 'nodes per second' and so on.
It reminded me just now of the old man in Squid Game - enormous smirk on his face in the Red Light Green Light opener ...
Why did he make the concession of 'well node just refers to 'consider' '  ?
To keep up the spiel he shouldn't have made that concession.
But looks like he's not going to concede much as to what 'consider' means.
This is all very relevant to the subject of 'chess will never be solved'.
It certainly will become even less likely - not even weak solving - if whoever is pushing solution is reticent about the terms he's using !

Oh wait -  he isn't 'cured'.
Post 2358.  He's ignoring the warnings from the moderators.
There's a red button within that post ...

Note that the post number will change if somebody deletes something.

It's worth remembering that a discrete game of perfect information like chess has states which tell you everything relevant to the future course of the game. A "position" is just a lot of the information associated with a state. The inclusion of e.p. and castling in states is very simple, and the information needed for the 50 move rule is equally simple - just one number - but the information needed for the 3-fold repetition rule (or any other repetition rule) is pretty much the entire history of the game. Hence FEN omits only the latter, while a full PGN provides what is needed to derive it. Minimally, you can get by with a PGN starting at the last irreversible move (initiated using a FEN).

While such a FEN+PGN describes the full state, many of them describe the same state. All of them that traverse the same bag of positions since the last irreversible move, to be precise. Even with this reduction, tablebases of such states are way too large to be practical.

Regarding repetition - that can be regarded as a position to be solved from.
Regarding change of strategy because 2 fold repetition has occurred -
I pointed out that even the opening position could re-occur.
So does that mean that every single position has to be multiplied by two - in case it has occured once before ?
If so - so be it.  The computer might have to solve for that too.
I'm suggesting though - that that could be a secondary project.
Like adding castling to tablebases.  Since that appears to be So difficult.

Same with 50 move stuff.  Mostly ignore it. 
Handle less than 50 and over the default time as separate projects..
Solve each position as if its at the beginning of the 50.  Make reports.
In other words - a way to begin to make 'weak solving' feasible.
Could be distinguished from 'heuristics'.

Of course in a tablebase repetition can be safely ignored for a lot of purposes. This is because if the position is decisive, there is a distance to mate which continually decreases with optimal play, thus repetition cannot occur. If the position is drawn previous positions have occurred - this can only make a draw occur more quickly with optimal play. The same position with no previous positions taken into account will be drawn as well.

So repetition only makes a difference to value when previous play has been suboptimal. If there was a win and a player goofed, a repetition could have occured or could be necessary to head for a win. The previously inaccurate play could even make the win disappear because all paths to it would run into a repetition, thus drawing.

I believe @MARattigan has drawn attention to such things.

That could burn up an awful lot of computer time.
Previous play was 'suboptimal'.
Seems to be getting back into 'games' again.

I'm suggesting a simple 'algorithm' - checkmate is availble or it isn't - but the computer doesn't have to confine its 'report' to that.
It can say 'in less than 50 with no capture/pawn move - or in more than 50 but pawn move/capture available to restart the count - 
or will exceed 50 move limitation but checkmate available in non-algebraic number of moves - or beyond 50 not known because of default computer time allotment.

Increase the results - rather than having to go back over ...
if it has to go back over - then its starting to head up into those crazier numbers like 10 to the sixtieth or to the 80th or over 100th.
Or perhaps its to note 'sub-optimal play' ?
If its gravitating back to 'games' then
It could get so that the finite universe Big Bangers would argue that universal entropy would have happened first ....

Yes. To have an oracle for perfect play in games - strong solution - you would need to deal with the full state including all previous positions. So not just a 32 piece tablebase, but something that has a size that is a huge power of the size of that tablebase.

To achieve the optimal result - weak solution - you don't need this, as you always reduce the distance to mate when there is a win and when there is a draw you don't care about repetitions.

" The number of edges from a node is the number of legal moves, and the number of edges to a node is the number of legal moves leading to the positions (or the number of legal reverse moves)."
Progress on 'nodes'.

"I hypothesise there is a legal sequence visiting every legal position (with just two kings) once (like in the even simpler knight's tour problem)"
Isn't that one obvious ?
There's always going to be many many legal sequences leading to each and every possible position with two Kings only.
The pieces are simply exchanged off.
But maybe that's not what you had in mind.  @Elroch.

'Equivalence classes' looks like a good term.
Groups of positions separated from each other by a capture or a pawn move.
Obviously only one such with 2 Kings and only one with 2 Kings plus any non-pawn piece.
With 2 Kings plus a pawn there would be at most - six.

Semi-strong still isn't applicable.  Just come up with some other nomenclature for it.  

As for attacking, let's not pretend you don't make attacks on people (individuals and the rest of humanity whose IQs are <160 in general)  constantly...

P.S. The "if you don't agree with my argument/definition, you are mindless" tactic is one of your weakest to date.  It's schoolyard stuff.

No - incorrrect in fact.

The game terminates when any such position occurs.

I think what @Elroch had in mind was a game with the dead position rule excised (and "position" with the special meaning assigned in art. 9.2)  when it's not obvious.

This statement by 'the guy' pathetically contradicts itself -
"It's very wrong to make comments about Elroch in that way. That's definitely about him. I wasn't happy that he told me I should be scientific but I've changed my mindset now and accept him for who he is. A finite universe banger."
Does 'the guy' even think he's serious?  Sometimes probably.
Appears to be a transient and cyclical situation.

We already know it terminates.
I think @Elroch was probably referring to sequences of moves arriving at different positions of the two Kings -
obviously there are many such sequences - but that's Too obvious ...
suggesting to me he had something else in mind. 
Involving that - but different from it.
Perhaps he'll clear it up.
Martin - we probably agree on this - regardless of what 'art 9.2' is -
no need to argue about that on which we agree - where we agree.


And you've confirmed it seems that the nature of strategy and 'prognosis' of positions might change according to how many moves from the 50 a position might be -
but I've suggested there's a workaround for that.
Rather than having to multiply each position to produce 100 different positions according to how far from the 50.
If that's going to be instituted - then maybe it could be reconciled with 'equivalence class'.  But that's not the workaround I'm suggesting.
A 100 x multiplier could be a big deal ?  Elroch seemed to argue maybe not - by arguing that the 'advantage' for the side needing to draw would increase.  It seems.
But just in terms of the load on the computers - 
if the whole project was going to take 100 years (crassly unlikely it would be that short) - then 100x takes it up to 10,000 years.
Are 100  years and 10,000 amounting to the same thing - as far as 'soon' is concerned ? 
Well 10K years could be regarded as in the same 'ballpark' as ten trillion years.
Whereas with 100 years - there could be computer improvements along the way. 
Greatly reducing it.   
So 100 years looks very much like 'will be solved' territory.
Downside:  No way to solve in 100 years or even in 100 million years !
Will they still have Tropicana orange juice then ?

#2364

"the information needed for the 50 move rule is equally simple - just one number"
++ That is right, but not needed.
In practice the 50 moves rule is never invoked before the 7-men endgame table base is reached. In the ICCF world championship there are 16 decisive games, and 108 draws: 81 by agreement, 17 by 3-fold repetition, 10 by reaching a 7-men endgame table base draw, none at all by the 50 moves rule. A rule that does not get invoked can be considered unwritten.

The high draw rate signifies that several of these games are likely ideal games with optimal moves. More draws means more games with an even number of errors than decisive games with an odd number of errors. The only logical explanation for that is that there are drawn games with 0 errors, less decisive games with 1 error, less drawn games with 2 errors, less decisive games with 3 errors...
The trend is even stronger after eliminating the games of the last player, who apparently had to resign games for extraneous reasons, maybe health problems.
https://www.iccf.com/event?id=85042 

"the information needed for the 3-fold repetition rule (or any other repetition rule) is pretty much the entire history of the game."
++ Yes, that is right, but easy to do. 
An average game lasts 39 moves with a standard deviation of 14.
It is easy and fast to compare the new FEN to the 39 previous FEN. 

"While such a FEN+PGN describes the full state"
++ No: a PGN describes all of its FEN; an array of FEN describes a PGN.
They are dual representations of the same.
Either the PGN or the array of FEN are enough.

Hypthesis 6. An error is not only what change a draw in a loss or a win in a draw, as you stated before. A more severe error can change a win in a loss. So if, let's say, Black makes a mistake and the game change from drawn to lost, but White makes a blunder, the game can turn into a win for Black, thus you can have a decisive game with an even number of errors. The same reasoning applies to drawn games, which may contain an odd number of errors.