# [Retros] MDR / postulates / hierarchy

Rol, Guus G.A.Rol at umcutrecht.nl
Mon Mar 3 13:30:53 EST 2008

Hi Roberto,

I haven't had the chance yet to look at your examples but I can give an off-hand reaction to your comments. It's actually not important whether or not the position or partial moves are "legal" for my approach, the only requirement is that they are "decidable". By that I mean that there is a way to decide, by means of your "hierarchical categories" or by my "dominant assumptions" or by "any other means", how to reduce the unknowns in the diagram position. As an example let's have Finzer's 6 solutions problem where a Queen on b5 could be underway to b2, but could also just be at b5 at the beginning of a full move. If there is no way to decide one way or the other, the problem is unsolvable. Finzer 'solved' it inappropriately by his stipulation "mate in less than a move". It doesn't work that way in chess logic since "asking for a mate in less than a move" does not imply that there must be a legal solution in less than a move! The correct stipulation should have been "Finzer half moves" allowed; fastest mate?" In this form he would first make his particular brand of half-moves ("Finzer half moves") part of the diagram presentation rules, give them equal rights to full-move assumptions, and only then issue the assignment. Because he didn't do that, compositions like these end up as humorous stage acts. One can laugh about it, but do not try to reason about it for that will lead nowhere.

There is enormous power in the proof game approach. As an example I will take your "underway illegal moves" category. The most effective way to construct proof games for these is not by finding the "last legal position" but "the next legal position"! Let's look for a proof game that is either "legal" or "legal when fixed by MDR-correction moves". Further definition: an MDR-correction is any FIDE-regulated correction to an MDR-position passing through your hierarchical category filter. Actually these two defintions state that (1) only legal and MDR-positions are allowed (2) there must be a prescription to decide that a position is of the MDR-type (3) there must be a FIDE-prescription to decide how the MDR-position can be returned into the legal framework (4) the evaluation hierarchy eliminates the less preferred MDR-assumptions.

With the above requirements in place, it is possible (but not always desirable) to have all MDR-categories co-exist, without need to specify the type in the stipulations. But of course every solver must have that same fore-knowledge contained in the previous paragraphs, otherwise he will fail in his analytical efforts. Conventions are an effective way to deliver such information as extension of orthodox chess. Another way is to define the "MDR" fairy type for all coexisting MDR-types and define the rules for the whole fairy type. And it doesn't really matter if we talk about dominant or recessive assumptions, or about hierarchies, since the objective of both is to reduce the MDR options and fill in the missing game state information.

Note: I doubt that there is a systemic definition of what is and what is not a minimum deviation from the rules. Examples: (1) lifting a piece but instead of dropping it somewhere, we press the clock (2) moving Ra7 to a8 and Queening the rook, forgetting it was not a pawn (3) completing an illegal move, following it up with a legal one and only then discovering the error (which must now be corrected) (4) touching a piece but not moving it. If any of these is permitted it will cook quite a lot of other potential MDR-problems, e.g. number 4 will cook the Finzer problem: since I have already completed the first part of a move - touching the piece - it can now be defended that the remaining part is "less than 1 move". My point is that all the MDR-types probably require list-definitions; a minimum deviation from the rules is only a minimal deviation from the rules if it appears on the list of predefined minimal deviations from the rules. Without such lists no MDR-problem is solvable unless by telepathic agreement.

Just some thoughts.

Best regards,
Guus Rol

-----Oorspronkelijk bericht-----
Van: retros-bounces at janko.at [mailto:retros-bounces at janko.at] Namens raosorio at fibertel.com.ar
Verzonden: donderdag 28 februari 2008 14:49
Aan: retros at janko.at
Onderwerp: [Retros] MDR / postulates / hierarchy

Hi Guus,

Your postulates are very clear and out of question dealing with legal positions (I think that the attempt to link to a posteriori - or to retro-variant logic is not just a dream but a matter of further work).

Regarding MDR, it has to be noted that we are dealing with "deviations" from the article 3. The MDR task is to build a general frame to contain in the same convention the cases of illegal full moves, underway legal moves and even underway illegal moves.
This is like having all the X-men in the same room and trying to keep order.

Look at this example,

Jorge Lois, Roberto Osorio Sergio Orce & José Verdejo Die Schwalbe 13071
8/8/2Rp4/4b2B/8/1p1p4/8/K1k2N2
#1 (RV?) Both kings under check

This problem would be now published as "#1 (RV) MDR Romantic Type". Problem intention is to show a last illegal full move (c4xd3 e.p.) and, under the MDR logic, apply the article 4.3 consequences (touched pieces; which one first?, motivating the RV, mate by white & mate by black).

But the MDR convention also contains the underway move cases (as in the twinning Example in my feb 27th message). Without some hierarchy, the problem would be cooked by a half legal move variant (a knight on c3, where white or black would be in the process to capture it).

For this reason, the MDR definition establishes that the correct solution is the one showing the highest preference according to the following order, -A full legal move -A full illegal move -The less incomplete legal move -The less incomplete illegal move

On top of this, MDR is a convention for retros and therefore all the genre conventions apply (no moves after mate, DR, etc). This keeps the X-men quite controlled. The "less incomplete" detail produces some fine situations. For instance,

8/8/8/8/8/5K2/p7/R4kN1
Mate? Circe - MDR

solved by Andrew Buchanan but that time the hierarchy had not been presented and therefore the discussion couldn't be complete.
It's not a full delivered mate because the position can't result from a full move.
It has to be noted that in Circe a capture is composed by three parts: to remove the opponent's piece, to move the own piece to the capture square and to install the captured piece on its homesquare. Candidates to be the underway move (legal) are KxBe1 / Rx(Q,R,B,N) on b1 to d1, In the process to deliver mate / b3xBa2 (after Bb1-
a2+) / b3xNa2 (after Nc1-a2+).
Excepting b3xBa2 and b3xNa2, all the rest show just 1 part of the move already done. On the contrary, b3xBa2 and b3xNa2 show 2 parts already done. But b3xBa2 would be an after mate move, so the correct underway retraction is b3xNa2.

Best,
Roberto Osorio

--------------------------------------------------------------------------------------------------
Hi Roberto,

Your pretty formal approach to MDR hooks in quite nicely to the system I have developed for the formalization of "missing game state information" conventions. Look at these postulates:

(1) A diagram is preferably viewed as a position at the start/end of a full move. I call that the "dominant assumption". The dominant assumption is the one you must assume in play if it is legal in at least one proofgame for the position.

(2) A diagram can be viewed as a position with a move underway only if inevitable. I call that the "recessive assumption". No proof game may exist where the diagram coincides with the completion or start of a full move.

Of course, (1) and (2) are the opposite sides of the same coin and are only separated for clarity. The first one corresponds in form to the castling convention, the second one to the e.p. convention. Evidently, the standard forms for these extended conventions can only be derived from your premise that "the MDR definition ALLOWS a deviation but not forces it" which is not true for some of the Finzer categories. With the 2 new genre postulates, no further stipulations are needed to identify an "MDR-problem" amidst "full-move problems" which is ideal from the design viewpoint.

A further advantage of this postulate form is that it may be seamlessly tied in to aposteriori logic - where a fractional move assumption could overtake a full move assumption if justified a posteriori - or to retro-variant logic - with partial solutions linked to different fractional retro-moves. But this paragraph is probably just my dream!

Guus Rol.

-----Oorspronkelijk bericht-----
Van: retros-bounces at janko.at [mailto:retros-bounces at janko.at] Namens raosorio at fibertel.com.ar
Verzonden: woensdag 27 februari 2008 14:16
Aan: retros at janko.at
Onderwerp: [Retros] MDR / promotion / fractional moves

Hi Mario,

Many thanks for this very valuable mail. This is the type of help we need indeed:
Antecedents. We knew the BG (Berührt - Geführt = Touched-Moved) one but not the article you are commenting here.

Any additional help would be very welcome.

Yes,. I got the Finzer's problem from the PDB.

The MDR convention we are working on attempts to establish a general frame for all the retroanalyzable types of deviation from article 3 (Laws of Chess). These deviations are classified as,

- Opportunity: leaving its own king in check; using already lost rights
- Completeness: the move does not complete all it's parts.
- Sequence: the move does not respect the legal order of its partial actions (promotion and castling)

We use "retroanalyzable" because the geometry of the move (complete or not) respects the article 3. On the other hand, the MDR definition ALLOWS a deviation but not forces it. Then, an MDR problem has to build a proper scenario for a correct solution that makes the situation "retroanalyzable".

We identified 16 deviation cases for the orthodox chess. These include cases resulting from a complete and illegal move (as in the BG problems) and from a fractional move (legal or illegal, as in the Finzer's article). We established a simple hierarchy in the scale completeness- legality that makes possible all these types to coexist under the same convention with a clear criteria to determine the correct solution.

An example,

Jorge Lois & Roberto Osorio
5bBR/p1pppPpp/Pp4p1/1N6/8/P7/kPP1P3/1RK5
Is white forced to deliver mate? MDR
b) wR from h8 to a8

A retro looking to the position shows that white is in the middle of a capturing process.
(other way, black side would be retrostalemate).

a) f7xRe8=... is being done, legal but not complete (1 part done, 2 undone). Then, according to article 4.6 the move has to be completed and it's mate despite of the type of piece it were promoted.

b) f7xRg8=... is being done, illegal and incomplete (2 parts done, 2 undone, illegal sequence). Then, a strange retraction (article 7.4) has to be made: to recover the position with the black rook on g8, but after that white is forced to capture this rook anyway (article 4.3, the rook was "touched"). To capture the rook white has to play f7xRg8=... but it's not committed at all to promote a bishop. So, the answer is no, since white is able to promote a knight.

This example illustrates my original point of "legal order of partial actions". A bishop promotion was clearly insinuated (half done) but there are no penalties. This is the most subtle case: a just formal illegality showing retroanalyzable traces.

Best,
Roberto Osorio