Hi Danny,
On Sun, 11 Nov 2012 23:16:06 -0000
"dannyjones183" <dannyjones183_at_yahoo.com> wrote:
> Rambling:
>
> I have a distant recollection of Shlomi Fish saying that FCS had problems
> finding a solution if he limited the number of moves.
You are right, but that's not precisely accurate. What I did say was that I
tried to limit the depth of the state in the solution, but given that I use
meta-moves and apply the raymond_prune, each step in the depth can consist of
several distinct atomic or column moves.
> I don't recall him ever
> saying that he'd been able to resolve that issue.
I have not tried to resolve it recently, and not sure how well the max_depth
feature is working. From what I recall, the problem was that the DFS scan found
states that were reachable in lower depths, but where found later and as a
result blocked the shorter solutions as well.
> Since then, he's introduced
> numerous additional filters, but he still has one deal in the first 32,000
> that doesn't solve in under 175 moves.
That *was* true. However, you need to understand that different configurations
of Freecell Solver yield different results. Right now with the latest
configuration ( Presets/testing-presets/mfi-with-2-more-scans.sh ) the solution
of the 175 moves board, and that of one other very long solution board have
been fixed, and the longest solution (FC-Pro-moves-wise) now stands at 92
moves). That's not too good either, but still much better.
The histogram of the solutions is now:
23 1
24 1
25 6
26 2
27 8
28 12
29 8
30 16
31 18
32 46
33 47
34 57
35 107
36 134
37 157
38 213
39 271
40 324
41 435
42 479
43 636
44 734
45 866
46 1027
47 1212
48 1244
49 1345
50 1516
51 1583
52 1655
53 1729
54 1623
55 1685
56 1610
57 1489
58 1430
59 1338
60 1139
61 1023
62 797
63 722
64 669
65 542
66 458
67 351
68 284
69 201
70 157
71 144
72 101
73 80
74 66
75 49
76 41
77 27
78 23
79 20
80 10
81 7
82 3
83 6
84 5
85 1
86 2
87 2
88 4
92 1
> If he ever wants to generate a useful
> solution for this deal, then I think it's time to admit that something is
> wrong and try to track it down. I'm willing to offer my assistance.
>
Any assistance for improving Freecell Solver is welcome. You can find the
source code for it here:
https://bitbucket.org/shlomif/fc-solve and submit
issues and/or pull requests.
> What's the number of the deal that forcing FCS to generate 175 moves? What's
> the FCS solution for this deal?
>
The number of the deal of 175 moves was 6,240 and its solution was:
[[Num Iters]]=197926
[[Num FCS Moves]]=206
[[Num FCPro Moves]]=175
[[Start]]
4a 4b 1c 14 45 8d 83 2h 63 a8
54 13 6a c6 45 61 85 58 54 6c
61 84 a6 8a 1h d8 68 6d 45 b6
2b 26 42 64 24 26 d2 27 46 56
87 8d c8 18 12 1c 17 54 82 51
b5 64 8b d8 78 75 85 8d b8 7b
78 7h 71 b7 46 81 d8 58 48 43
57 5b 5d 2h 6h 6h 2h 2h 6h 6h
56 45 c4 34 6c 16 1h 12 1h 15
43 d1 b1 53 8d 8h 71 54 a5 17
1b 1a d1 23 2d b2 1b d1 72 6d
a6 27 26 c2 34 6a 6c a6 2a c2
8c 8h 3h 87 8h c8 31 5c 62 b5
a6 13 72 7b 31 8a b7 c8 13 1b
27 a1 5c 72 d5 6a 43 b6 5d 46
c5 36 d4 1b 3c 3h 7h 27 b1 4d
72 3b 3h c4 1c 31 3h c3 4c b1
6b 6h c4 32 3c
[[End]]
Reached Board No. 6240 at 1352570506.213502 (total_num_iters=2367638)
=====
The number of the deal with the 169 moves solution is #31,465 and its solution
is:
[[Num Iters]]=187082
[[Num FCS Moves]]=203
[[Num FCPro Moves]]=169
[[Start]]
6a 6b 1c 16 18 1d 81 a2 83 18
1a b1 d8 6b 6d 61 a6 16 d1 b1
4a 4d 41 4b 48 7h 6h 6h b6 4b
c4 b4 7c 27 42 4b a4 84 83 43
4a b4 24 2b 25 21 a2 71 7h 32
57 38 28 23 23 2a b2 a2 5b 5a
52 d2 56 c5 b6 4d 4c a4 5b d5
72 7h 7a b7 5d c5 4b 17 13 d5
84v5 28 25 8c 8d 81 85 8h 3h 5h
b8 c5 2b 2c d2 52 1d a1 6a 6h
34 3h 1h b1 43 25 8b d8 1d a1
2a b2 8b c8 1c d1 7d 7h 18 61
6h b6 73 c7 8b 6c a6 8a 56 b8
2b a2 1a 41 4h 82 d8 1d a1 7a
b7 47 65 c4 2b 74 7c b2 56 3b
3h 6h c7 6c 47 d6 36 1d a1 8a
74 34 37 64 b8 3b 3h 3h 32
[[End]]
Reached Board No. 31465 at 1352572587.711962 (total_num_iters=3819882)
I'm not sure how usable these "standard notation" solutions are so I'm also
attaching the fc-solve -p -t -sam solutions, which are more regular. You can
find a parser, processor and verifier for them here:
http://fc-solve.shlomifish.org/verify-code/
> In addition, I can supply a solution that can be used to debug FCS and learn
> where/why it refuses to implement my solution. I'll save specific suggestions
> about performing the debug until after I find out if Shlomi Fish is
> interested in outside assistance.
I always am.
Regards,
Shlomi Fish
--
-----------------------------------------------------------------
Shlomi Fish http://www.shlomifish.org/
Parody of "The Fountainhead" - http://shlom.in/towtf
He who reinvents the wheel, will understand much better how a wheel works.
Please reply to list if it's a mailing list post - http://shlom.in/reply .
- application/x-ygp-stripped attachment: stored
Received on Sun Nov 11 2012 - 22:27:05 IST