The New Sudoku Players' Forum

[Mauricio]

Thanks for the compliments, now a 4x4x4x4 Jelly:

Code: Select all

+-------+-------+-------+
| . . . | . . . | . . . |
| . . . | . 1 . | . . . |
| . 1 2 | 3 . 4 | 5 6 . |
+-------+-------+-------+
| . . . | . . . | . . . |
| . 2 3 | . . . | 7 8 . |
| . 4 7 | . 6 . | 1 2 . |
+-------+-------+-------+
| . . . | . . . | . . . |
| . 3 1 | 8 . 7 | 6 4 . |
| . 5 8 | . . . | 2 3 . |
+-------+-------+-------+   ER = 9.2 with Jellyfish disabled


Code: Select all

.............1.....123.456...........23...78..47.6.12...........318.764..58...23.


My ultimate goal (for now) is to generate a minimal symmetric puzzle such that a Jellyfish is the only technique that can be applied (between the set of techniques that tarek asks, for instance) and solves the puzzle.

[tarek]

Code: Select all

.............1.....123.456...........23...78..47.6.12...........318.764..58...23.


My ultimate goal (for now) is to generate a minimal symmetric puzzle such that a Jellyfish is the only technique that can be applied (between the set of techniques that tarek asks, for instance) and solves the puzzle.

amazing, a living breathing 4x4x4x4 Jellyfish which scores an 8.

Code: Select all

.............1.....123.456...........23...78..47.6.12...........318.764..58...23.#017#Mauricio-002#8#1

tarek

[Mauricio]

This one is probably the most "extreme" (SE wise) that I found with that pattern:

Code: Select all

.............1.....123.456...........27...38..36.5.42...........682.715..74...83.


It should score an 8 too. ER is 9.5 if Jellyfish is disabled.

My search was pattern based. I gotta find another good pattern to start a new search.

[tarek]

This one is probably the most "extreme" (SE wise) that I found with that pattern:

Code: Select all

.............1.....123.456...........27...38..36.5.42...........682.715..74...83.


It should score an 8 too. ER is 9.5 if Jellyfish is disabled

true

Code: Select all

.............1.....123.456...........27...38..36.5.42...........682.715..74...83.#018#Mauricio-003#8#1

tarek

[Pat]

#Puzzle#Chronology#ID#Score

.......63..763.48...3..8..1.3...41.5.1..8..3.5.63...7.2..7..3...74.562..65.......#008#daj95376-003#4.2

this puzzle can be solved with just 1 Jellyfish ( + "singles" )

and therefore should be considered pure Jellyfish even by strict rules -- very pure -- in my view

[tarek]

#Puzzle#Chronology#ID#Score

.......63..763.48...3..8..1.3...41.5.1..8..3.5.63...7.2..7..3...74.562..65.......#008#daj95376-003#4.2

this puzzle can be solved with just 1 Jellyfish ( + "singles" )

and therefore should be considered pure Jellyfish even by strict rules -- very pure -- in my view

The last modification to the rules & rating methods in the head post, shows that puzzles acquire a score, that score is used to categorise the puzzle at question.
"Pure Jellyfish Extra Lenient" category has to score at lest a 4, the puzzle in question scores 4.2. So yes it is a pure Jellyfish in the Extra lenient category.

tarek

[Mauricio]

I think this one needs 2 Jellyfishes

Code: Select all

.............1.....123.456...........35...78..81.2.35...........57...63..638.721.


Add: Not exactly, if you apply the second Jellyfish first, then singles cascade and the other Jellyfish dissapears, how is this classified?

[tarek]

.............1.....123.456...........35...78..81.2.35...........57...63..638.721.

it scores an 8
if all Jellyfish seen AT ONE TIME are considered as 1 (Batch) then the number of batches is 1 ......... that is the number reported when I say JF >=1
However The total possible is 2 Jellyfish .... Manual solving or exhaustive electronic search (can be done) can confirm if both are needed or not.

The idea is for both the numbers above to stay stable after applying the lenient set of techniqes (which it does)....

similar puzzles (tarek-006 & daj95376-003)

stranger still, daj95376-004 has an increase in the total possible Jellyfish from 1 to 2 when we go beyond singles. (the extra one probably providing a red herring which is good but also may provide solution)...examples of these are plenty on the swordfish collection

tarek

tarek

[gsf]

to mirror the changes in the rules for the swordfish collection, here are the updated rules[code]Rules:
1. Valid vanilla 9*9 sudoku
2. NEEDS a Jellyfish to solve after applying the following set of techniques:
2.1. Singles (Hidden/Naked)
2.2. Box-Line/Line-Box interactions
2.3. Subsets (doubles,triples,quads) (Hidden/Naked)
2.4. X-Wing
2.5. Swordfish
...
Score interpretaion:
Score = 1 ==> Invalid puzzle
Score = 2 ==> Requires other than a Jellyfish to solve
Score > 2 ==> Requires a Jellyfish to solve

I'm trying to code this up and am confused by (2.) (for starters)
what about the techniques allowable after the jellyfish is used?
also, can you add something so that score==2 means only one condition, not two?

[tarek]

to mirror the changes in the rules for the swordfish collection, here are the updated rules[code]Rules:
1. Valid vanilla 9*9 sudoku
2. NEEDS a Jellyfish to solve after applying the following set of techniques:
2.1. Singles (Hidden/Naked)
2.2. Box-Line/Line-Box interactions
2.3. Subsets (doubles,triples,quads) (Hidden/Naked)
2.4. X-Wing
2.5. Swordfish
...
Score interpretaion:
Score = 1 ==> Invalid puzzle
Score = 2 ==> Requires other than a Jellyfish to solve
Score > 2 ==> Requires a Jellyfish to solve

I'm trying to code this up and am confused by (2.) (for starters)
what about the techniques allowable after the jellyfish is used?
also, can you add something so that score==2 means only one condition, not two?

I thought this was one of my best attempts of being clear

Point 2 essentially means :
solvable using (2.1+2.2+2.3+2.4+2.5)+Jellyfish
Not solvable using (2.1+2.2+2.3+2.4+2.5) alone

A puzzles that fails this will not go through the rest of the checks and would score a "2"

The scores interpretation can be clearer but essentially ">" does not point at the score but actually means "more than". You maybe quoting an earlier post. but the latest rules & score interpretations are found in the head post of this thread.

Point 4 has to do with the number of Jellyfish
4.1 deals with Jellyfish Batches
4.2 deals with Total Jellyfish POSSIBLY encounterd during batch solution.

Point 7 is personal preference to see if some simpler techniques can bypass the Jellyfish. any puzzle that goes through this check must have already passed all previous 6 checks & therefore should score a 7 if it fails or 8 if it passes.

here are some puzzles to explain the scoring system:
[code]#Puzzle#Chronology#ID#Score#JF>=
.1.....2.34.....56...2.6.....76285.....5.7.....69342.....4.5...56.....84.7.....1.#013#JPF-001#2#0
1.........2.345.....5...4...6.7...3..8..1..5..1...9.4...7...5.....283.6.........8#010#TTHsieh-001#3#0
7...2......1..9.6..8.4..2....6....5.9.......1.3....7....2..6.9..4.7..8......3...6#001#tarek-001#4.1#0
.......63..763.48...3..8..1.3...41.5.1..8..3.5.63...7.2..7..3...74.562..65.......#008#daj95376-003#4.2#0
4..2.7.8...3..85.6.786...4.2.....61.....2.....17.....4.5...346.6.15..2...3.1.2..8#009#daj95376-004#6#1
...197....4.....7...9...3..8...3...27..9.2..43...6...1..2...4...1.....9....524...#004#tarek-004#7#1
.............1.....123.456...........27...38..36.5.42...........682.715..74...83.#018#Mauricio-003#8#1[/code]

tarek

[gsf]

thanks for the explanations tarek

I'm getting a clearer picture of a generalization for this and related threads

select a target technique (jellyfish here) and divide the solution into three phases, all in batch mode:
(1) the techniques applied before the target technique
(2) the target technique (1 or more batches)
(3) the techniques applied after the first target technique, leading to the solution

puzzles can then be categorized by the technique counts in (1) (2) (3)
as a first stab we can just do a SE-like count of the "hardest" technique applied in (1) (2) (3)

puzzles can be additionally categorized by minimality and symmetry

to experiment with this I updated my solver 2008-01-11 to handle top level constraint (technique) grouping
the grouping is left-to-right: once a group is exhausted it is discarded for the remainder of the solution
along with this is a way to test and extract the counts within the groups after a puzzle is solved

for this thread try these options to experiment:

Code: Select all

-BG -q'{FNBTHW2W3}:W4:{FNBTHWXY}'


: separates the three constraint groups
guessing (G) is disabled
constraints up to swordfish inclusive are applied until there's no progress
then all jellyfish (W4) are exhausted
then the remaining constraints are applied until the puzzle is solved
(XY cover most uniqueness/bug, but no direct correspondence to SE techniques)

here's the format that extracts the group stats, symmetry and minimality info
its a mouthful

Code: Select all

-f'%#0v # %2#in %3(score)f |  %2(h1)x %2(b1)x %2(i1)x    %2(h2)x %2(b2)x %2(i2)x    %2(h3)x %2(b3)x %2(i3)x  |  %(S!=0)x %(C1)[0][2][1]x'


score is one of the fields in my input data -- it will be null non-annotaed puzzle input
h1 is the highest constraint applied index from group 1, counting from 1 on the left within each group
(-q constraints may expand, use -T0x10 to correlate indices with the expanded groups
the actual correlation is not as important as higher index => harder technique)
b1 is the number of batches in group 1 containing the highest constraint
i1is the total number of instances of the highest constraint in group 1
similarly for groups 2 and 3
the second last column is 0:no-symmetry 1:symmetric
and the last column is 0:minimal 1:symmetric-minimal 2:non-minimal

for the posted puzzle # original-ordinal score

Code: Select all

400207080003008506078600040200000610000020000017000004050003460601500200030102008 #  9   6
700020000001009060080400200006000050900000001030000700002006090040700800000030006 #  1 4.1
050000070200000006000456000007030400006501200009070100000812000900000005010000020 #  5   7
600070000001005030020400600003000010900000002040000700009003050070600900000010003 #  3   8
700060000003001080020700600008000040100000003090000200006008010050200700000070008 #  2   8
100000000020345000005000400060700030080010050010009040007000500000283060000000008 # 10   3
040028071870600020000000000004800260001040900086002400000000000060009043130270090 #  6   7
200003580000020001000405600000904068005010900610802000001508000300040000026300004 #  7   7
010000020340000056000604000005786200000905000007423500000307000960000032050000010 # 15   2
003004009000070050200600800009000004040000030800000700004003002020060000700800600 # 11   8
000000010000809257020000900082901045047080190190403720001000080938105000050000000 # 12   6
010000020340000056000206000007628500000507000006934200000405000560000084070000010 # 13   2
000000063007630480003008001030004105010080030506300070200700300074056200650000000 #  8 4.2
000197000040000070009000300800030002700902004300060001002000400010000090000524000 #  4   7
010000020230000045000502000006374800000906000004258600000407000740000038050000010 # 14   2


here are the stats sorted by cols 3 - 13

Code: Select all

#  9   6 |   3  1  1     1  1  2     1 15 15  |  1 0
#  1 4.1 |   3  1  1     1  1  2     1 15 15  |  1 2
#  5   7 |   3  2  3     1  1  2     2  1  1  |  1 2
#  3   8 |   4  1  1     1  1  1     2  4  4  |  1 2
#  2   8 |   4  1  1     1  1  2     2  2  2  |  1 2
# 10   3 |   4  2  3     1  1  2     2  1  1  |  1 2
#  6   7 |   5  1  1     1  1  1     2  1  1  |  1 1
#  7   7 |   5  1  1     1  1  2     2  2  2  |  1 1
# 15   2 |   5  1  1     1  1  2    15  1  1  |  1 2
# 11   8 |   5  1  1     1  1  3     2  1  1  |  1 1
# 12   6 |   6  1  1     1  1  2     1 14 14  |  1 0
# 13   2 |   6  1  2     1  1  2    15  1  1  |  1 2
#  8 4.2 |  11  1  1     1  1  2     1  7  7  |  1 0
#  4   7 |  12  1  1     1  1  2     2  1  1  |  1 1
# 14   2 |  12  1  2     1  1  2    15  1  1  |  1 2


I hope this makes sense, and I know it doesn't match the scores
but does it capture the nuances you are looking for?

[tarek]

for this thread try these options to experiment:

Code: Select all

-BG -q'{FNBTHW2W3}:W4:{FNBTHWXY}'


.
.
.
I hope this makes sense, and I know it doesn't match the scores
but does it capture the nuances you are looking for?

hmmmm, I'm sure this implemntation has many applications other than this thread.

However let put some remarks...
essentially for this thread: your second bracketed contstraints should have only FN used by pure Jellyfish puzzles (your h3 column should be <=2) there are 3 puzzles having 15 (those that correspond to my score 2)....

if you add a -QNFW4-G then h1 should show if this is a benchmark puzzles if (h1<=2) or if it contains red Herrings (the bigger the better )

The Second group captures 1/2 of my point 4 however it should match a set similar data if your first bracketed constraints were NF alone

the second bracketed constraints has many applications also but not much apart from what I mentiened above ...

so the formula above (for the purpose of this thread) is missing 2 things
1. a -QNFW4-G check
2. a -QNFW4-G (W4) data (batching & total instances) (%2(b1)x %2(i1)x)

The reson for this is to insure that a manual solver cannot encounter more Jellyfish following FN than after using FNBHW2W3

%2(b1)x following FN <= %2(b1)x following FNBHW2W3 (my point 4.1)
if above is "=" then %2(i1)x following FN <= %2(i1)x following FNBHW2W3 (my point 4.2)

My point 7 of the rules can be translated like this.....

Code: Select all

In the abscence of W4, how far up the heirarchy of your techniques can you rech before finding a solution

...It deosen't have to do very much with the previous 6 points apart form subjecting the Lenient Pure Jellyfish to its torment.

tarek

[daj95376]

Say, I bet you boys have a side-line job writing users' manuals for newbies to IBM, Oracle, and Microsoft software

[tarek]

I have a better idea

Code: Select all

 -q'{FNBTHWXY}:W4

Would collect data as follows:

1. Lowest rank heirarchy technique reached that would solve if followed by W4 (should be 1 or 2 for pure jellyfish)
2. highest rank heirarchy technique reached where in the abscence of W4 there would be no solution without guessing (should be at least W3)
3. highest rank heirarchy technique actually used (this would show benchmark vs. red herring status)

Data collected along the way regarding batches of W4 & total instances of W4 would then be used also.

tarek

[Pat]

I think this one needs 2 Jellyfishes

Code: Select all

.............1.....123.456...........35...78..81.2.35...........57...63..638.721.



Add: Not exactly, if you apply the second Jellyfish first, then singles cascade and the other Jellyfish disappears, how is this classified?

just 1 Jellyfish was needed when intermediate methods were allowed

-qNFBHTW-G

now need 2 Jellyfish merely because intermediate methods are disabled

-qNFW4-G

of course i accept it ( under my most-lenient rules );
but this is exactly the situation i expect tarek to reject, not-very-pure