Adrian, your mailer does not seem to have cut the long lines in this
message. Please fix its configuration for next time.
On Sat, 6 Jul 2002, Adrian Ettlinger wrote:
> Shlomi, a while back, commented that about 100 messages had passed
> between he and I while working on integrating Freecell Solver into
> FcPro. Now, after another 100, we're finally at the point where it can be
> semi-publicly distributed. You can download it from
> http://www.rrhistorical.com/rrdata/Fcpro65/ .
>
I'll put a link to it on the Freecell Solver homepage.
> This version incorporates Shlomi's recently-released version 2.4.3.
> This fixed up a bug which was uncovered in our testing -- a very weird
> one causing a crash in Windows on certain intermediate positions.
>
It was not particularily weird - I just forgot to initialize a few things.
Sometimes, different systems can pre-initialize a variable to different
values which may lead to unexpected results.
> We are awaiting release of upcoming FCS version 2.6, which will
> implement atomic moves, to eliminate false impossibles.
Actually, version 2.6 can run either with atomic moves, or without them.
Use of atomic moves seems to make it considerably slower.
> We ran a test
> in which we ran all positions in all the solutions which are posted on
> Mike Keller's website -- these are human-generated solutions to roughly
> 500 deals from the M/S original 32,000. In testing approximately 25,000
> positions, there were 185 false impossibles, a ratio of 1 to 135. In
> actuality, false impossibles are more prone to occur in the middle part
> of a solution (never, in this sample, from an original position), so the
> ratio is no doubt higher for positions likely to interest most players.
> We are looking forward to replacing version 2.4.3 with 2.6, which
> hopefully will eliminate all false impossibles.
>
I suggest we give the users a choice between several presets some of them
would be based on atomic moves and some of them would be based on
meta-moves.
> Comparing the speed of FCS for range solving in the 32,000 M/S Freecell
> positions, FCS now leads in the speed competition, being roughly twice
> as fast as both the FcPro solver and Patsolve.
Are you sure you checked it against Patsolve's speedy mode? I know that
is a very fast mode of operation. In any case, the development version of
Freecell Solver, with the computer generated solving presets, can solve
the MS 32,000 in roughly three times the speed of the best preset I could
find for 2.4.x.
> This applies to the
> standard four freecell game. At two and one freecells, it is relatively
> much less effective, but Shlomi has not yet made any efforts to optimize
> it for that application.
>
I'll see about it sometimes later.
> We said "semi-public" distribution above because we have no plans of
> our own to publicize this in any other arena except for this forum. We
> will leave it to others to spread the word as they see fit.
Like I said, I'll put it on my web-site. However, I do suggest that
Michael Keller's site would carry the most up-to-date version.
Regards.
Shlomi Fish
>
> ---------------------------Adrian Ettlinger
>
>
----------------------------------------------------------------------
Shlomi Fish shlomif_at_vipe.technion.ac.il
Home Page:
http://t2.technion.ac.il/~shlomif/
Home E-mail: shlomif_at_iglu.org.il
He who re-invents the wheel, understands much better how a wheel works.
Received on Sat Jul 06 2002 - 22:57:36 IDT