[Retros] Pawn to-the-8th-rank mates - response to Francois Labelle's work
flab at EECS.Berkeley.EDU
Wed Dec 27 02:56:27 EST 2006
Andrew Buchanan wrote:
> Thanks for all this wonderful work.
Thank you. As you might suspect, adapting my program for this task was
relatively easy, the bulk of the work is in the postprocessing.
> There is one area where alas we appear to disagree. To my mind, the way
> the task was specified, there was no mention of minimizing the number of
> units captured.
I don't think that the way we keep track of records was specified at all.
For example for the task "King mated on the 7th/8th rank" people learned
that the number of captures was a criterion because one day it appeared in
Hirokaz's record list. For the task "Pawn to-the-8th-rank mate" maybe that
criterion is in force but it just so happens that new records never
required it? We can't tell.
So maybe it's up to the record keeper to decide.
Or maybe it's time we use a standard convention. Simple tasks like this
one appear regularly on the mailing list and it would be nice to have a
standard way to keep track of records so that the person who initiated a
challenge doesn't have to mention it, and the competing composers don't
have to ask.
Personally I don't care whether the number of captures is a criterion or
not, I will just program whatever people prefer.
Another question is whether we should keep track of slower opposite color
games or not. Personally I think that we should not, because a PG in n
moves can often be modified to give a PG in n+0.5 moves (this is not
guaranteed, but neither can we guarantee a PG in n+1.0 moves). Again, this
is open to debate.
> Subjective criteria that I then tried to apply where possible (within
> the very limited design space) were in no particular order [...]
Some of these criteria could be used objectively. For record-keeping, the
question is which ones are important enough to trump the "timestamp"
> There is a broader question then. If the grand database enthusiasm for
> retro problems is going to take off, then we will need to record in the
> database, not just all the positions, but also the evaluation criteria
> used, and be able to associate with each position the values that the
> criteria take.
There are really two things: the "task" (= constraint) and the "evaluation
method" (= judge). If we can agree on a standard evaluation method for use
on the retros mailing list, then only the task will have to be described
for most of the mailing list problems.
> In terms of the original question: "What is a shortest PG for a given
> mate?", we now have an answer for 56 of the 88 cases. Yay!
Does anyone find the number of cases excessive? Personally I think that
only one record for each different type of mate is worth archiving
(dividing the number of proof games by 22). Of course if you guys are
having fun I won't stop your party, but you see where this is going: a
list of proof games ending with every possible move.
More information about the Retros