[Retros] Re: Massacre SPGs

[Francois Labelle]

Andrew Buchanan wrote:


> Interesting. Well, we'll see. The figures of 300 & 90 are presumably

> just fudge factors. How do they relate to the factor of 50 times the

> number of operations you estimate for x=6?


Yes, 200 and 90 are just fudge factors, and 50 is just a guess for what
comes next after 200 and 90, knowing that the series tends to 1. It's
based on no other information, and I'm not really interested in discussing
a better estimate unless you claim that it could be outside the interval
[20,80].


> If you have the space, you would presumably always want to use it to the

> max. My only point was that if space runs out, you can take advantage of

> the fact that the problem is basically decomposable. How do you see the

> time increasing as you increase the number of batches?


I hope I won't have to find out! But to get a feeling for what would
happen I watched in more detail what happened from ply 20 to ply 21:

- The number of positions got multiplied by 2.10.
- Then the position merging step multiplied it by 0.62.

This gives a net factor of 1.30 from ply 20 to ply 21 (which you can also
compute from my table on the web). So transposition detection (aka
position merging) appears to be quite important.


> Or maybe this could be distributed across many computers. However there

> would be a lot of work involved to set that up.


I think that this is the kind of setup you'd see when people compute
something important like the lift of a proposed airplane wing as part of a
multi-million-dollar project. To answer a recreational problem it's hardly
justified.

Hopefully I'll find a (1+1) SPG at x=5, and the case x=6 won't really
matter anymore. :)


> > Ironically there is no SPG with the kings on their home squares

> > yet (no wKe1, bKe8 for x=3,4).

>

> I am confused by this, since it doesn't seem to square with what you

> write later...


Sorry, replace "SPG" with "(1+1) SPG with possibly many solutions" and
reread. I thought this was the context when I wrote it, but I should have
been more explicit. From now on I'll definitely write "SPG with possibly
many solutions" each time I mean that, or maybe I'll just say "diagram"
since any legal diagram is an SPG with possibly many solutions.


> Some/many chess problem solvers, myself included, don't really enjoy

> solving these kinds of problems,


I guess some of these problems may lack in theme or paradox, but some of
them like my "5... Rh1#, game?" are more of the why-didn't-I-think-of-that
type.

I told you about 800 at home SPGs in 4.0..14.0 moves. That's a wide range
of difficulty level to choose from. I was wondering if you would claim
that those computer-composed at home SPGs will in general be of
significantly lower "quality" than the human-composed ones. Imagine a
blind experiment where the evaluators don't know who composed what, in
case they're biased against computer creations.

If they're of the same quality, then my computer is simply the most
prolific at home SPG composer on Earth, and people may want to put their
effort elsewhere if they feel that it's now pointless to compete for that
particular class of chess composition.


> but I do enjoy marvelling that these positions exist.


Then my post to the mailing list should be sufficient. Example: I told you
that an at-home SPG with only 6 units "exists". The exact diagram isn't
important. :)


> That's why I think an article in a magazine which gives some of the

> background to the work is better than just submitting the problems to a

> magazine.


Ok, well I think that I should wait for x=5 to be complete, since it's now
clear that I can get it done.


> > - stalemate in 15.5 (4+1)


I was wondering, is this the first known stalemate SPG?

As always, thanks for your feedback.

	Francois