[Retros] Cooking a famous AP-problem?

liskov at im.bas-net.by liskov at im.bas-net.by
Fri Jun 29 19:33:18 EDT 2007

Dear retro-friends,

I see no reason to revise the conventional treatment of AP
after Petrovic and to reinterpret the AP-study by Rinder.
In my opinion, extra arguments by Guus (and the application of
the PF-logic in the way that Guus proposes) are not convincing.
It is entirely B's responsibility to justify the (attempted)
forced e.p. key, W may well counteract him, and B fails unless
he manages to castle in all the lines (not manages by ANY reason:
destroyed castling, perpetuum checks, a premature mate, etc.).
Since the move R8a7+(??) destroys castling, it cannot help B to
justify e.p. and, thus, any further game after it is senseless
wrt to the key justification (regardless of a triple repetition,
a draw or not, etc.). [Here "??" marks illegal moves, including
ones that definitely make the whole line illegal.] Executing e.p.
restricts only B, not W; it's only a (forced) attempt to move.

Now two remarks apropos, on 50 moves.

I too don't know any AP-problem (type Petrovic) based on
the 50-moves rule. In the following problem
(V.Liskovets and A.Kornilov, 1981, "Schwalbe", h.69)
4k2r/4ppp1/2S5/2P5/1PpP3P/P1P1P1Pp/p4K1p/7R (11+9)
B-> (AP) Can Black win?
we succeeded only partially: a blockade of B's castling is
sufficient to disqualify the e.p. key; however, unfortunately,
W can further destroy castling.

As to automatic draw, let me quote once more (we discussed this
in January) Art.5 of Piran's Codex 1958:
"If the solution necessarily asks for more than 50 moves without
any capture, pawn-move or castling, the position is not a draw.
If provable by retroanalysis that one or more of such moves
directly preceded the solution and the total amounts to or
surmounts 50 moves, the position is automatically a draw."

Valery Liskovets

Rol, Guus <G.A.Rol at umcutrecht.nl> wrote:

> Yesterday my brain must have been short-circuited for a while trying to

> sell you standard "AP type Petrovic" for the move turnover type "AP

> after Keym". Fortunately this doesn't affect the crux of the story but

> I'll correct it here for the sake of clarity.


> Guus Rol.


> (correction of yesterdays entry)


> Dear retro-friends,


> Another chapter in my discussion of retro-strategic principles

> is this famous problem by Gert Rinder. Now, I haven't voiced my concerns

> on the foundations of AP-logic here (yet) and I won't be doing so at

> this point. I'd rather focus on 2 aspects of this composition:


> (1) the post factum principle in the move choice for AP problems


> (2) how can the automatic repetition rule cook this problem - to

> create an even better one!


> Author: Gert Rinder

> Position: r2bk3/p1Rpp3/P1p2p2/KpP2P2/1P2p3/1P6/p7/8

> Author stipulation: Draw (AP type Petrovic)

> My stipulation: Black wins!


> Starting from the second aspect I suggest this solution: 1.cxb6

> e.p. (requiring justification) axb6+ 2.Kxb6 Pa1R 3.Kb7 R1xa6 4.Rc8

> (first time) R8a7+!! 5.Kb8 Ra8+ 6.Kb7 (second time) R8a7+ 7.Kb8 Ra8+

> 8.Kb7 (third time!); if you know PB-R309, you know what comes next. The

> position can't be considered a draw since there is no proof that the

> "same" position occurred thrice! Continuing play however - by whatever

> move, e.g. R6a7 mate - proves prior black castling right absolutely, and

> thereby proves the e.p. right a posteriori.


> Now there are several ways to argue against this approach - we

> don't want this automatic draw convention, it didn't exist at the time,

> this is a strange mix of AP and PF (not a good objection), etc - but all

> objections overlook the angle of having just created a beautiful new

> solution to a famous AP-problem. Not only does it contain all of the

> content of the previous one - the original solution becoming the try -

> but it adds a new element as well which is not readily available in most

> AP-compositions. I couldn't find another existing AP-problem cooked

> through this procedure, but I could find some that will be easily cooked

> by the 50-moves rule if it ever became rationalized and automatic. It is

> a characteristic trait of AP-problems that short AP-solutions can be

> affirmed or refuted by long "virtual" variants.


> Now back to point (1). In one of my previous posts I introduced

> the (rather arbitrary) concept of "game enforcers". This is a good case

> for demontrating what I mean by that. The principle behind game

> enforcement is that "a PF-solution continues under most conditions

> unless can be shown that it stops". One of the reasons is that PF

> implies that all choices from all "acceptable pasts" are available

> simultaneously. White may castle long if he can, he may castle short if

> he can - his choice is arbitrary even if the castlings are mutually

> exclusive. In Gert Rinders problem the question is the choice between 2

> pasts: (I) a past in which white is stalemated in the diagram (II) a

> past in which e.p. is legal and can be justified a posteriori. PF logic

> says he can chose either one of them and therefore must chose the one

> that will leave him a move. Stalemate is the absence of even a single

> option to move. PF-logic shows that the stalemate option need not be

> accounted for by retro-variants. So indeed, white starts and black wins

> in this composition.


> Note: I use the common descriptor PF here, but I'd rather use PA

> (Post Actum) for its major component. Left for another post.


> Guus Rol.

More information about the Retros mailing list