Website founded by
Milan Velimirović
in 2006

9:08 UTC
ISC 2021



Remember me

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

Rating lists

MatPlus.Net Forum General Case study for "Golden Age" keyword
You can only view this page!
(1) Posted by Andrew Buchanan [Saturday, Jun 5, 2021 19:09]

Case study for "Golden Age" keyword

This is a *long* response to Joose Norri's comments in another thread: ( If it doesn't interest you, don't read it. It's something I'd wanted to write for a while: a case study of how the "Golden Age" keyword can allow us to resolve conflicts of rule interpretation by:
(1) protecting our past heritage of problems that were sound in their time, and the context in which they were created,
(2) allowing composers today to compose in those past modes.
(3) giving us a clear focus on today's rules and conventions for the majority of compositions.

This is a foundational meta-model for the Fairy Chess Classification programme, and happens to use as a worked example the Dead Position Rule (DP) which Joose is interested in. If Golden Age can handle this, it can handle anything.


Thanks Joose, I really appreciate your response. I tend to take rules of all kinds quite literally (although I am not interested in word games or tricks except as occasional jokes). I am going to explain things carefully but once only. This is not going to become a long debate.

My goal is to make the FIDE Rules and Codex Conventions properly welcoming to fairy chess, who have been excluded from orthodoxy for too long. I am not a fairy expert, but I enjoy it very much, and note that fairy chess essentially includes all of mathematics, and the fairy classification is not going to work unless we are rigorous about the orthodox rules and conventions. Although there are some rules and conventions that I am much fonder of than others, I can't allow this subjectivity to interfere. And logic is sometimes not enough - judgement or interpretation are required. But we can try to identify those jumps and ask if, on balance, the responses are reasonable. Here I try to align myself with not just the letter of the documents, but what I assess to be the spirit behind them.

Please point out any holes carefully. Are you with me so far?


I've changed my mind on I realized I wasn't following my own guidelines - oops, it's been a while since I was in this part of the forest! I'm classing this problem as "Golden Age", not unsound. What on earth does that mean?


The sequence of events was: 1997 "Yerevan" Laws, 1998 Wilts/Geissler, 1999 Codex *adopting* the Yerevan Laws. So the problem was sound when it was published. But even if it was published in 2000, it would still be ok. My approach is one of "maximal soundness" - a problem deserves to be considered charitably to be the soundest it was at any evolving stage of the rules/conventions - potentially even *before* the problem was composed if it is "retrospective" in nature! E.g. many modern problems based on Dummy Pawns (legal in 1862) are classed as "Golden Age".

This approach does something absolutely vital: it allows the rules and conventions to evolve, which they certainly weren't before. We were stuck in a logjam.

Another huge win is it allows individual composers to differ in opinion. "Good fences make good neighbours" is an essential motto for co-existing independent spirits! Joose, if you compose a retro problem which goes against the DP rule (and you have a perfect right to do so), then it can be classified as Golden Age, pre-1999, and it will be sound forever.

I hope to work with Stewart Reuben, who's writing a history of the rules of chess, to build a timeline for all significantly different versions. Then it will be much easier for problems to be assigned their natural place in history. This will resolve for example the current mess on the 50M convention (stuck on whether castling zeros the move counter, would you believe?), and issues with castling & e.p. for 3Rep. This is an important part of the anthropological rigour that we need to bring in to our hobby to track the context of historical problems, which otherwise lose their meaning and significance.

This can be extended to classify according to schools of composition, themes of tourneys, conditions etc. We have to be the curators of our own rich cultural history, and proper classification by context is a key anthropological principle. To be clear, we don't have to relegate inconsistent ideas to the fairy zone or to the past, with just one winner retaining orthodoxy.

By the same token, the two Labelle problems composed in 2004 are fair-and-square sound. They were sound when composed and remain so today. Joose, we have to respect the existing Laws and Conventions where they are clear, and orthodox problems constructed by fellow independent spirits. Otherwise we can't hope to move to the next level of logical rigour. We can't build a proper foundation for mathematics (by which I mean all of fairy chess) unless we have a clear idea for each chess problem what set of rules & conventions it applies to. That doesn't stop you making your own compositions of course, as I will discuss later.


First let's clear up a distraction. DR (Dead Reckoning) was a term that I personally introduced to suggest where the DP rule has surprising effects both retro and forward. But it doesn't add enough and I think I will stop using it right now, forever. Sorry for confusion. Let's just talk of the DP rule.

Dead Position rule (DP) has been in the FIDE Laws since 1997 as a replacement and generalization of the earlier rule of draw by insufficient material, which was deemed unsatisfactory. Through many new versions of the Laws in the intervening 24 years, this has remained unchanged, and my discussions with Arbiters suggest that it is viewed as a success. Hurray!

Simply: if the position is such that neither player can possibly checkmate the opponent’s king, the game is drawn. In common with checkmate and stalemate, the rule also requires that the final move be legal. This final point is extremely interesting and suggestive in itself, but that's for another time.

The Codex of 1999 adopted Articles 1-5 of the Yerevan Laws, which included the DP rule. In 2015, a group of studyists realized that this rule threatened to break a few of their studies. I was overjoyed that these folk accepted that there was a real issue, so I supported them 100%, working with Kjell Widlert and Michel Caillaud to put the DP rule into a similar status to the 50M rule (although without the aforementioned 50M mess, which is specific to 50M). So DP rule, like 50M rule, would apply only to retro problems.

Some DP problems were broken that day, and not all proved fixable. I am especially sorry to Ronald Turnbull, who had developed a number of excellent forward DP problems, and am grateful for his acceptance. However, these problems do live on in their own Golden Age.

Explicitly on that day, I concurred with Kjell that "retro" itself does not have a fixed definition. In that, I am a fan of the philosopher Wittgenstein, who said we cannot define exactly words such "love", "game" etc. Fortunately, there's a lot of case law in this area. We'll come onto that in a minute.

Final change, with no impact on DP: in 2019, the Codex was updated to follow the better-written Laws of 2018.

So bottom line: DP rule applies by default today to all retro problems.


Another distraction we need to clear away: this time on your side, Joose.

There are certainly some similarities between help play and proof games, the most important of which is that both genres (or genre and sub-genre to be strict) are non-adversarial. But the obvious difference is that there is choice in the end position in a help play. One form is a string fixed at one end only, while the other a string is fixed at both ends. As a result, help plays are mostly much shorter, almost always ending in mate, and the very long ones have a single line of play (rather becoming a jigsaw of moves by many pieces that needs to be rendered total in a unique way). One way to attack a long helpmate is indeed to enumerate the possible end positions, and then tackle them one by one as an A-B proof game. However, artistically the two stipulations have very different design centres despite the minor overlap, as the WFCC recognizes in their distinct genre classifications.

But the slight resemblance to helpmates is all irrelevant because even if *every* single PG was classified as both retros *and* helpmates, the DP rule would still apply to PGs because they are more retro than just about anything else in chess problems.


As I stated above, there is no fixed definition of retro. So we accumulate fibres to make the ship's cable, in Wittgenstein's beautiful metaphor. (1) PGs are part of the Retro genre for Album purposes. (2) There is a list of stipulations that WinChloe considers retro, including PGs. (3) All PGs are classed as retro in PDB. (4) Most tellingly, retros are defined at the Retro Corner website as "problems where position legality is a key element". Let's build on that. The Codex states: "A position is legal/illegal if it can/cannot be reached by a sequence of moves from the initial array" (and there is a similar definition in the Rules). So finding a proof game for a position is about the most retro thing one can do! A proof game in this pure sense can’t be unique, so we are asked in a PG to find the *shortest* proof game out of those which are available. The notion of shortest, most elegant proof, “the Proof from the Book”, is a not just a mathematical subject of interest, and has deep connections to computability. In practice, we are often *given* the number of moves as a hint (and so a potential solver can judge the difficulty of the task they are presented with, before diving in), but in most cases it’s equivalent to looking for the SPG: the Proof from the Book. So all PGs are retros.


>Just to mention, I do not agree that P0005399 is rendered 'no solution' by the DR rule. This is help play, with the diagram position as the target, the concept of checkmate is >meaningless - unless you try to reach the diagram via a game that enters and leaves checkmate.

As shown, help analogy is tangential and irrelevant to applicability of DP. It's all about retro.

>Logically, in PGs DR means any move that makes it impossible to reach the target position. So if you target the basic position of the Ruy Lopez, then 1.d4 is DR. Every solver uses >this even without knowing about DR...

No this is a fallacious straw man argument, and an incorrect usage of the word "logically". I think you mean "jokingly". DP has nothing to do the stipulation or with Ruy Lopez. Let's stick to the rules & conventions please.

>But DR can be used in SPGs as an additional rule, at least theoretically. You can think of a SPG which has two solutions without DR, but with it, only one, because the other is DR >too early.

Yes absolutely can have one orthodox twin, and another "Golden Age" one without DP! A non-PG retro which did this is Classification of mixed mode twins is always a problem. Methodologically, I tend to think that each twin gets its own classification. And then the overall problem inherits any keyword or genre of any subpart. Do you agree?

> Ah, I overlooked the last Labelle. Well, I do not agree. I consider it correct only with the addition 'DR in effect' or something similar.

Well, you would be completely wrong, Joose, and small dogs would laugh at you as you walk down the street :-) Francois' *two* problems are already 100% correct already, as per Rules & Codex, without any graffiti in the stipulation. The composer may opt to add a courtesy hint about DP - preferably including that word "hint" so it doesn't get mistaken for a fairy condition! But in articles and databases there's usually enough context from the text or keywords to provide enough of a hint. For problems which are *not* retro however, I am always scrupulous to modify the stipulation, even though this is also Golden Age, e.g. So really belt-and-braces here.

Please feel free to compose your own problems which go against the DP rule: I think it's more accurate to classify them as Golden Age rather than fairy, because there is no new condition, or board or pieces, just reverting to an earlier rule set. Let's keep "fairy" better focused.

Or just accept that DP is just an exotic but real part of chess: it's nice that the rules throw up a few peculiar things. It adds to the spice of it all. Yes?

Now, let's get away from this sideline, to look at all those other game-ending PGs that I took the trouble to list. You're welcome, Joose! Thanks for your time.
(Read Only)pid=21208
(2) Posted by Joose Norri [Sunday, Jun 6, 2021 20:14]

Just a simple question (to start with, perhaps?). Are you saying that the problem 'Kh6 Bg6 - Kh8 Bg8, helpstalemate in one, two solutions' in fact has no solution?
(Read Only)pid=21220
(3) Posted by Andrew Buchanan [Monday, Jun 7, 2021 03:07]

Hi Joose. This is not a retro problem so under Codex Art 17A, by default the DP rule does not apply. There are two solutions
(Read Only)pid=21224
(4) Posted by Hauke Reddmann [Monday, Jun 7, 2021 09:42]

Uh-oh, Trollman Hauke just saw another glaring problem
that can't turn up in a OTB game - draw is draw,
regardless of stalemate, dead, 75 or fivefold.
As long as your stipulation says =, still no problems.
But you can imagine a position where the finishing
move fulfils all of the four, but specifically one
(most common: stalemate) is stipulated. I would default
to "don't bother, stalemate+dead is still stalemate",
especially as stalemate is a subset of dead by definition,
but problemist, always fond for teh lulz, might explicitely
define otherwise.

So, let's take Joost's position, the stipulation being
"shortest helpdraw, but not (not even also) by dead".

Try 1: 50 moves
Try 2: Threefold repetition
Solution (ROT13): Oynpx bssref n qenj. Juvgr npprcgf.
(exit left, madly giggling)
(Read Only)pid=21227
(5) Posted by Joost de Heer [Monday, Jun 7, 2021 11:15]

I remember a Caillaud composition from Strategems, it's neither in PDB nor in WinChloe, as far as I can see. Stipulation was h'='2. Solution was 1. A X 2. -A -X. One could prove that the previous moves were -1. -A -X and the only legal moves following the second move were 3. A X, so there was an inevitable draw by 3rd repetition after the second move.

EDIT: Didn't search properly in WinChloe. The solution was slightly different than I remembered too.
(= 16+4 )

Michel Caillaud, Strategems 2001

Last moves were -1. Rc6-a6(1) Kc8-b7 -2. Bc7-d6
1. Kc8 Rc6 2. Kb7 (2) Rc5!! '='
The position is a forced draw because the next few (forced moves) are 3. Kb6 Rc6 4. Kb7 (3) with a draw by 3rd repetition.
(Read Only)pid=21228
(6) Posted by Andrew Buchanan [Tuesday, Jun 8, 2021 03:57]

Thanks for your great replies, all.

My main theme was "Golden Age". However I didn't emphasize enough another idea offering prevention rather than cure. The current situation is a "zero-sum" competition where one lucky interpretation wins the prize of orthodoxy, and all others are cast out into the darkness of effective heresy (whatever the Codex preamble consoles them with).

But really in many cases, there are parallel models which are all ok. There are numerous ways to endow a pawn on the first rank with movement capabilities. There may be knock-on implications for en passant, or presumed retro history for a pawn on the second rank. At this stage in the mess, one can't say that any one of these interpretations has primacy over the others. Although a particular condition like Einstein may include just one of the interpretations in its bundle. So as long as a problem is clear about the model it assumes, there's no issue, and I think the fairy world in particular will include many examples of this.

There are many aspects where it's the right answer to have a single orthodox model, but we should be aware that this approach is potentially pernicious and divisive, and should not be applied blindly without reflecting on the impact to all Stakeholders (particularly composers past, present & future, and solvers present & future).

There are three areas where the rules must be adapted when moving from OTB to composition:

(a) No game history. We start just with a position, and game state may need to be added, assumed or derived.

(b) Removing tournament machinery. The FIDE Laws do not separate cleanly into rules of the game and tournament guidelines, although an attempt is made. Physical board, wood, hands, clocks, arbiters, score-sheet, etc all must be abstracted away. This involves simplifying away certain protocols like 3Rep and (what has not yet been done) 50M. The most important simplification here (except for jokes) is that each move is atomic. It moves from one position to the next along the graph. Prior to any move in the problem there is a static analysis.

(c) Automating human decisions. The stipulation does the job of saying what the players will attempt to do. Are they acting together or in competition? But there are other decisions which need to be elided away: resignation could defeat all self-mates were it allowed, draw offers could provide short solutions to any helpstalemate. It's obvious that these have no part. The decision-making in claims is another piece. Separate from the history of the game, it needs to be clarified whether the player would always claim, and what happens if one is looking retrospectively at a position where someone *didn't* claim although they could.

(3) PAT & DP
Hauke: as usual you raise numerous trenchant questions. I never see you as a troll. I think you meant Joose not Joost.

Years ago, I used to think that dead implies stalemate and vice versa. But that's clearly not the intention of the Laws. Now I think I get it.

Say White has the move over the board. How many legal moves are there? If zero, game over, immediately. Check means mate, else pat. Otherwise, there is at least one legal move, and we look ahead in the game tree to find any helpmate for either side. If none, then game over, by DP. Otherwise the game is living, and White can play. So pat and DP are fundamentally different. I commend this approach to you. (I have ignored 3Rep, 5Rep, 50M & 75M here.)

One area where the rules/conventions do not opine is the degree to which DP rule sees the conventions. We are simply not told, and this area needs two different models, neither supreme over the other, as both interpretations seem to be able to support beautiful problems. This is a complicated area, and really needs the full "film of the book" adaptation perspective. As a soundbite I would say it beneficial for DP rule to see 3Rep/50M convention, but counter-productive to interact with the castling/e.p. conventions. There's also the different question of whether the castling/e.p. conventions see the DP rule. But let's not go down that rabbit hole now.

It's certainly true that 3Rep & 50M can coincide: I made a problem which had that feature a few weeks ago. However, since 3Rep involves repetition, this can never combine with pat. And interaction with DP rule depends what conventions DP sees, not just in 3Rep/50M but in other aspects of state such as castling & e.p. rights.

(5) Joost: Michel Caillaud's beautiful problem was already in PDB at
(Read Only)pid=21232
(7) Posted by Kevin Begley [Thursday, Jun 17, 2021 09:33]

I like the "golden age" idea.
Pardon me, you may have addressed this, and I missed it in skimming, but what about problems that were uncertain when published?
Are they to be considered "silver age" ?
(Read Only)pid=21270

No more posts

MatPlus.Net Forum General Case study for "Golden Age" keyword