[Retros] Last move Type D

[Mario Richter]

Hi Andrew,

you wrote:


> (1) One issue with Type D is that there are ludicrously many possible

tasks.

> One area of focus could be positions where the White last move & the Black

> one are the same, e.g.:

>    P-B/P-B/10 K1krB3/p1pp1p2/8/8/8/8/6P1/7b


That issue (White last move & the Black one are the same) was what I meant
with:


>> Can at least all Type D records be found, where Black's

>>  and White's last move is of the same nature?


Btw., I think that many combinations are definitively not possible.
(e.g. I suppose, that KxQ/* is not realizable).
My estimation is, that only approx. 300 combinations are possible.



> (2) I've only looked at a few of these, but there is one obvious

translation

> from German needed in B--/B--.


I was afraid of that - my original notices were in german, so I checked
my mail twice - and nevertheless overloked one. (It seems, that this was
the only typo). But the FENs should be consulted anyway ... :-)


> (3) We need a canonical ordering, to see easily what's been done already.

> Alphabetical sorting is fine, but then there is an issue in ordering the

WTM

> & BTM solutions.


I suggest we do it the same way as Type A records are ordered, i.e. not
alphabetically, but based on the piece value (K>Q>R>B>N>P).


> ... there is an issue in ordering the WTM & BTM solutions. I think in

> the canonical form, these should be alphabetically ordered as well

> (e.g. K-/N- not N-/K-) no matter which side plays which.

> Unless there is a position which can only be realized one way round...


I thought about such a possibility - but the only idea I had was
that a parity argument might play a role ...


Thank you for your comments

  mario