Website founded by
Milan Velimirović
in 2006

2:17 UTC
ISC 2020



Remember me

Forgot your
Click here!
to create your account if you don't already have one.

Rating lists


MatPlus.Net Forum General The Retro-Strategy convention: what is the correct way to interpret it?
You can only view this page!
Page: [Previous] [Next] 1 2
(21) Posted by Guus Rol [Tuesday, Dec 8, 2020 19:38]

@Bojan_Basic: Just saw your link to the Mat Plus problem 2014. Unfortunately I don't know know what the fairy condition "Disparate" means. Can you explain? Why is there no white king? Anything to do with "Disparate"?

May I guess? You can't play the piece type your opponent played his last move with.
(Read Only)pid=20254
(22) Posted by Bojan Basic [Tuesday, Dec 8, 2020 20:19]

 QUOTE'P0000758%27 is my ear reddening past. In the year (I think 2006) before I completed my retro-theory I placed this comment in the PDB and it is WRONG

I have to apologize, it seems that I did not pay enough attention to this problem, and got it completely wrong. It seemed to me that the point was something like this: White has some possible moves, but since it cannot be proved which particular one of them is legal in the given position, he is not entitled to play any of them, and thus he is stalemated (this is why I wrote that the same issue occurs in another two problems I supplied)—if this were the case, I could not agree with that interpretation. But this is not at all what the problem is about: White maybe has the e.p. capture, but no other moves beside it; since the e.p. capture cannot be proved (by the AP convention), we take that it is not possible. Therefore, we agree on this, sorry for the confusion caused. :)

May I guess? You can't play the piece type your opponent played his last move with.

Yes, this is the definition. (Absence of the wK does not have anything to do with this Disparate, I guess that it is absent simply because it is not relevant for the scheme, just imagine it on any random place on the board.) But the initial question about Disparate was (in my opinion) untangled very nicely by Andrew; I mentioned that topic because of the last page and the problems mentioned there, after which the discussion stopped, and for me these problems defy all the laws of logic (and the laws of chess themselves). Thank you for your comment, it seems that we agree on this point also.
(Read Only)pid=20255
(23) Posted by Guus Rol [Tuesday, Dec 8, 2020 20:42]

@Bojan_Basic: OK, let's leave it here. Issues on the retro-logics will return until they are at rest in everybody's mind!

One more comment. Issues such as for Liskovets retro-volage problem - if I get it right because part of the text is missing in the PDB - are neatly resolvable by a generic "elevation convention" - applicable to all similar situations as well. When all exit moves from a position are blocked by the conventions while it is provable that one of the exits must be open in all proof games, you should give them "elevated rights/privileges" (computer term). Or more appropriate here, "elevated permissions". Before you can do this you must "group" them such that you can prove that one of the exit moves in the group must be certainly legal. You might say that all these moves are now temporarily raised to the level of "castling permission". So instead of executing none of them, you can execute any of them - the player decides! Where there is no king, everyone is king! This may or may not lead to duals. Managing those is the composers task. Everything is better than being stuck in the great undecidable void.
(Read Only)pid=20257
(24) Posted by Andrew Buchanan [Wednesday, Dec 9, 2020 03:00]

We are all independent spirits, and I must say that for me one of the coolest, best, top features of RS is that through hard work one can construct positions in which no forward move at all is permitted, even if it's certain that in each particular proof game, there exists a legal move. This is a beautiful, funny, elegant, emergent property. I really like it. It's disrespectful to think that this is something which would need to be fixed, and it would be vandalistic to impose some complicated, complicating, cobbled-together, band-aid bodge such as is apparently being vaguely alluded to, as usual with insufficient specifics. No smilies this time. OK one, just to indicate that I'm relaxed, but no more than one because I'm not joking :)

This is not just an aesthetic position. Every architectural complication which is proposed to "improve" the conventions must interact with all the others, and make it easy and pleasant to develop and extend further. I see no evidence, based on the sketches presented, that a simpler picture would actually emerge. My intuition, based on a humble career of working in IT architecture, particularly for data & information, is that this would lead to a conceptual logjam, and the system of conventions would become progressively unmaintainable. Without proper specifics, it's impossible for the reader to assess whether everything could indeed work harmoniously together. The apparent lack of awareness in what I've seen so far of the risk of architectural gridlock in the proposals, or enough details for us to judge for ourselves, makes me skeptical, sorry.

How can we be sure, for example, that the so-called elevator convention interacts harmoniously with the so-called macro-move concept bandied about as the flavour of the day a few days ago?

There is a correlation between aesthetics and architectural fitness, although the former is more subjective than the latter. We all here love complexity, or perhaps it's more accurate to say we love the *resolution* of complexity. That repeated satisfaction has to be against a background of underlying relative simplicity, so that the complexity is genuine chess complexity, and to the greatest extent possible, the referee doesn't start playing football.

Work within PRA if you must, but please leave RS alone. Simple is good. Thanks for your time.
(Read Only)pid=20259
(25) Posted by Guus Rol [Wednesday, Dec 9, 2020 09:50]

Sorry, I can only laugh about these arguments. All conventions exist to supply substitutes for missing history information. Without these conventions all of our RS, AP and PRA problems would be undecidable just as Liskovets problem. And people would be sending one another messages boasting about the new undecidable diagrams they found - instead of having beautiful retro-problems. All the elevated convention does is supply a solution to an information hole that was overlooked.

Note that this convention would be a supplement to the RS- and AP-logics only; the situation can never happen in PRA. It is totally RS in nature because it leaves the move and history choice between equals to the player which is exactly what RS does.

You need not worry about the consequences, I did and thought them out over a decade ago. Never need to worry about solutions based on proof games. Only arguments without proof game backup are in danger of insanity.

You are not a scientist in character because you love the magic (of undecidability) over the science. My mission is different.
(Read Only)pid=20260
(26) Posted by Guus Rol [Wednesday, Dec 9, 2020 15:51]

The definitve annihilation of Liskovets paradoxical retro-volage problem:

#1 (Retro-volage, RS) (= 2+14 )

You can find the traditional comments in the PDB here:
Peculiarly the comments in the PDB appear to have been aborted at some random point. Disturbing but quite a good methaphor for the status of the whole composition when I think of it.

Let me start by saying there are a number of aesthetical and logical elements in this problem which are appealing! And those are precisely the causes of its failure. Whenever we have a good idea we want to believe it is presentable in some of the environments available to us. Orthodox composers quickly learn the reality lesson that not everything can be shown within the constraints of their chosen composition types. But those in the fairy domain are not that easily deterred. They will tweak their chess laws until they find a way to present their darlings. And that's where accidents start happening. Fairy types are often multi-dimensional. They integrate with the orthodox chess laws, they combine with other fairy types and they may try to incorporate retrograde logics to achieve a desired effect. But it is a complex bundle as a number of these components may not be suitable to seamlessly perform with the others. So there may be undecidabilities, even contradictions until they are ironed out. If they ever are.

I will show a direct contradiction in the reasoning on the Liskovets problem. First some facts:
(a) the retro volage game rules are a fairy layer over orthodox chess rules
(b) the retrologic (in casu RS or PF which is the same) is orthodox with one extra convention for the retro volage state.
(c) the retro-logics demand that al least one proof game exists for a legal diagram from the PAS. All definitions are based on that minimum.

This is not just theory because the authors reasoning with regard to the volage state (on or off) of the rooks completely depends on the available proof games with the possible histories for all the rooks. Without requirement for proof games, any rook could be in any state and no solution is possible - or perhaps 13 solutions which I would have to analyze.

Next something weird happens. White plays Rh5 (which gives check to the king and is conventionally OK) and somebody shouts: checkmate! Ordinarily he would be immediate manhandled out of the room but we have arrived in a very permissive environment. The former rigor with regard to the proof games for volage states is gone and replaced with sloppy negligence toward the past. Who cares there is no proof game for checkmate, we did that on the last move. Don't tell us we have to do that again!

Indeed, I tell you you do. There is a psychological cause for this error. Checkmate is commonly tested against the future and not against the past. Lord Saurons eye turns north and thereby fails to notice what goes wrong in the south.

In a pro forma attempt to go forward the solver is challenged to find the next move. The conventions won't allow him to play one and he decides to rest in the satisfaction of the checkmate. But that is an obvious contradiction with the certain conclusion after the move Rh5. Information conventions supplement retro-analysis but can't contradict it.

The justifiable conclusions at this stage would be:

(a) it is not checkmate - because of retro-analysis independent of conventions
(b) there is no move the conventions permit us to play.

Another way of phrasing these conclusions is: "the position after Rh5 is check, not mate and the next move is undecidable". If the composer is happy with that conclusion then he is welcome to proceed and advertise his creation. Though - it is clearly not mate in 1!

P.S. He could tackle the contradiction with the "elevated rights" convention I suggested but that wouldn't save his #1 setting without extensive tweaking.
(Read Only)pid=20263
(27) Posted by Andrew Buchanan [Wednesday, Dec 9, 2020 20:11]

I can see that I've needled you. I have a right to my own subjective preferences, as do you. And even a "scientist" can be ok with some things not being decidable within a system, and understands that their existence doesn't stop great work from being done elsewhere within that same system. Some holes are not worth fixing: it's not just the 80/20 rule, but I'm sure you will agree that a larger formal system will have its own holes, probably larger, and most importantly: Architectural Complexity is Death.

Let's remember: we are not talking about solving particle physics, we're talking about designing a framework for a puzzle/game. Mark Rosewater at Wizards of the Coast is the best expert I know at this:

Better to start with something architecturally simpler that people can use, and can extend gradually to match increased understanding by a wider group. The "private language" nature of your ideas seems to me a real bottleneck, frankly. This argues for keeping explanations very simple and grounded in real problems, as you are now doing in some posts. More fun too. As we've discussed elsewhere, I am open to that pathway, but please don't try to merge my work in until you have a clean actionable set of rules for bread-and-butter retro conventions. Stop getting ahead of yourself: keep it simple and focused. Thanks.
(Read Only)pid=20264
(28) Posted by Hauke Reddmann [Wednesday, Dec 9, 2020 20:47]

Before you two go on un- and reweaving the rainbow (egad, my
popcultural insinuations are so brilliant you need shades :-)
- I still call for a completely formalized retro system, so any
difference can be resolved a la mode Leibniz: Calculemus!
(Hey, it worked even for theology and pin-dancing angels -
after Llull invented his disks, no beef about theological
disputes ever broke out and all world lives in peace
hitherto :-)
(Read Only)pid=20266
(29) Posted by Guus Rol [Wednesday, Dec 9, 2020 22:55]

@Andrew: The issue of the Liskovets volage problem was raised by Bojan_Basic and not triggered by anything you wrote. Until of course you participated with your viewpoint. Then I looked it up and saw you commented before.

You appear to believe that the state of my theory is related to the bits I communicate to you. When you say I shouldn't get ahead of myself you suggest I am still moving in the development of that theory. Not so. The theory has been complete since 2007 and has not significantly expanded since that time. By our communication you get it bit by bit but I am not developing it in those parts. It was actually designed the same way I taught myself the Gödel formula - in one compact streak of innovation over a small time span. For instance, I didn't need to construct an argument against Liskovets "stuff". It all came from cash memory since I worked through similar constructions in mind experiments. There is nothing new or challenging here.

Undecidability is either the result of poor design, or bad reasoning, or essential incompleteness of the system. It's sometimes forgivable, sometimes not. Whatever is the case, there is a proper way to handle undecidability and it is not the way Liskovets did it. You cannot skip over it and pretend it did not happen and draw some random conclusion. It's an endpoint of sorts. You go back home and do something else and let nobody know you have ever been there. The time police is always watching - Hauke warned me. Btw, math permits you to solve undecidability in a higher level model which supplements missing info in a different way. You probably don't wonna go there I guess.

Universes are complex but not nearly as complex as when they collide. It's a mess of competing black holes. That's why my retro universe looks complex. Whenever you see it, it is in instant collision with yours. But in itself it is not that complicated or at least not more complicated than the problems posed to it. That's why I still believe it can be taught to eleven year olds - provided they need no deprogramming from an earlier religion :-)
(Read Only)pid=20267
(30) Posted by Guus Rol [Wednesday, Dec 9, 2020 23:07]

@Hauke: Are you sure you want a completely formalized retro-system? Did you know these edit out all citations of pop culture, time travel and especially Leibnitz? Chances are even your name sticks to the filter :-)
(Read Only)pid=20268
(31) Posted by Andrew Buchanan [Thursday, Dec 10, 2020 03:48]

I’ve got stuff I need to do today so I will just make two points I should have verbalized before. First, the word “undecidability” is not being used in a proper context. Just saying that the natural operation of an RS approach means that there are a few places where all moves are blocked. I think this is a cool contrast with PRA.

Second, if your stuff is available since 2007 and somehow it’s not magically communicated to your desired audience this indicates that you have a perceived complexity problem. So instead of blaming your audience for having “closed minds” I suggest you follow the Rosewater link. WotC have managed to sell a game with 20,000 different moving parts and a 200 page rule manual to millions of people - they are far from being a perfect organisation but do understand something about complexity.

Good luck!
(Read Only)pid=20270
(32) Posted by Guus Rol [Thursday, Dec 10, 2020 11:11]

The applicable term is "paradigmatic shift". It is easy to bring something new and explain it to a reasonably intelligent person. It's much different when the new thing requires to do away with the comfortable old thing. In fact, the intelligent person is more likely to be stubborn because he has invested that intelligence in creating a private image of reality. As you probably recall I wrote that 90% of my "book" would be about deprogramming the existing "language paradigm" and 10% about explaining my suggested "process paradigm". Everything in my system is algorithmizable in principle!

The process system is also much more powerful than the language system. I already know a way to turn Liskovets problem into something much better with the "elevated rights". It retains all the content of the current one - because like Liskovets you need to prove the conventional undecidability before invoking elevated rights - and adds extra logic. However, that version would suddenly be rational and provable and that would frustrate those who only like Liskovets creation because it is magical. Can't serve two masters.

Btw, In 2007 I decided to first publish the example problems before presenting the theory in the hope that intelligent people would pick up on the logic. I was skeptical of success until you told me (hurray!) one of my pivotal examples won 1st prize! If I am mad, I am not alone!
(Read Only)pid=20273

No more posts
Page: [Previous] [Next] 1 2

MatPlus.Net Forum General The Retro-Strategy convention: what is the correct way to interpret it?