blue wrote:I get a much smaller pattern count, after VPTs ... ~38e9.
Yes, the right number is 128447994798305000 / 3359232 = 38 237 309 837
blue wrote:I get a much smaller pattern count, after VPTs ... ~38e9.
dobrichev wrote:The search possibly could be improved by
- Checking whether "dead clues" entirely cover a known UA set
- Checking for redundant clues at earlier stages
- Partial reusing of the intermediate search results for "neghbour" grids - those that differ in permutation of digits within a single UA
dobrichev wrote:I played with dynamic addition of UA after unsuccessful check for a single solution and found it unproductive. Large UA sets don't add value.
dobrichev wrote:Yes, the right number is 128447994798305000 / 3359232 = 38 237 309 837
Serg wrote:...I calculated (manually) low bound for number of e-d 17-clue patterns. I got low bound 3.8e10 (if I was not wrong in calculations). It does not contradict with your exact number of 17-clue patterns - 5.5e10 (appox.)...
Serg wrote:There are 77 millions e-d 17-clue patterns, having 1 clue in one box and 2 clues per each remaining box (see the thread cited above). Though number of e-d 17-clue patterns, having 1 clue in one row and 2 clues per each remaining row, is probably different, but surely comparable with 77 millions.
champagne, are you ready to search through 77 millions of patterns? (Distribution 122222222 is not the only possible distribution for 17-clue patterns.)
Serg
222222221
3.......0
3......00
3.......1
4.......0
4......00
4.......1
5.......0
5.......1
6........
Serg wrote:There are 77 millions e-d 17-clue patterns, having 1 clue in one box and 2 clues per each remaining box (see the thread cited above). Though number of e-d 17-clue patterns, having 1 clue in one row and 2 clues per each remaining row, is probably different, but surely comparable with 77 millions.
champagne, are you ready to search through 77 millions of patterns? (Distribution 122222222 is not the only possible distribution for 17-clue patterns.)
Serg
coloin wrote:There appear to be eight known puzzles with row,column and box counts of 222222221
There appear to be sixteen puzzles known with just row and column counts of 222222221
Only 2 puzzles in these lists satisfy the minirow constraint
C
1111..1.. 1
1111..... 387
111...... 1847
11.11.1.. 26
11.11.... 1235
11.1..1.. 6537
11.1..... 39031
11....... 86
1..1..... 2
Serg wrote:Hi, champagne!
If my understanding is correct, you are planning to do exhaustive search for class of pattern having following properties (constraints):
1. Each pattern contains 17 clues.
2. One row of the pattern contains 1 clues, remaining rows contain 2 clues each.
3. Each minirow of the pattern contains not more than 1 clue.
4. You consider essentially different patterns only.
Let's call this class of patterns as "Little subset of 17-clue patterns set".
Let's calculate number of patterns in "Little subset" class.
Approximate lower bound for number of essentially different patterns in this class can be obtained by dividing total number of such patterns (not e-d) by number of VPT for this class.
Total number of patterns can be calculated in this way.
1. There are 9 variants to select row which will contain 1 clue.
2. There are 9 variants of row containing 1 clue.
3. There are 27 variants of row containing 2 clues.
So, total number of patterns is 9 x 9 x 27^8 = 2.28 x 10^13 (approx.)
Number of VPT for this class can be evaluated as 6^8 (all classic VPT, except of transposing, because transposing can destroy class properties).
Therefore, lower bound for number of essentially different patterns in "Little subset" class can be estimated as (2.28 x 10^13)/(1.68 x 10^6) = 1.38 x 10^7 = 13,800,000 patterns. Too many patterns for "Little subset" to my mind.
Serg
+---+---+---+
|1..|3..|...|
|.2.|...|.5.|
|...|..4|..6|
+---+---+---+
|7..|...|...|
|...|...|...|
|..8|...|...|
+---+---+---+
|...|...|...|
|.9.|...|...|
|..1|...|...|
+---+---+---+
..1..2..3.2..1..4.5..6..1....2..4..1.5..6..7.7..2..3....7..5..8.8..7..9.9..3..2.. # C27/S4.da/M1.16.3
2...4...6..4..2.3..7.8..1....24...1.1...3...2.3...54....5..9.7..1.6..2..6...7...9 # C27.M/S4.da/M1.31.2
..2..3.1..3..1.2..5..6....4..5..7.2..9..4.8..6..9....7.8..9..5.2..5..4....3..8..9 # C27/S2.d/M1.33.1
..1.2.3...2...1..43..5...6...42...8.9...3...2.3...67...4...3..75..8...4...2.6.1.. # C27.M/S2.p/M1.13.4
1....2..3.9..4..1...73..5....21....9.6..2..5.9....56....3..41...7..8..2.2..6....8 # C27.M/S4.da/M1.11.6
2...3.1....14...2..5...6..3.6.2..4..3...5..7...7..3..65..8..7...9..1..4...6..2..9 # C27/S2.d
.2..7...84..6..9....5..4.7..1.2..3..2...1...4..3..5.6..9.5..6....1..7..37...3..5. # C27.M/S4.da/M1.16.1
1..3..2...2..4..1...3..6..52...8..7..7.5..3....9..1..67...9.5...4.1....8..8..2.6. # C27/S2.d
..2.1.3...1...2.4.5..3....2..5..7.8.6...3...4.4.6..9..7....4..6.3.2...5...4.9.7.. # C27.M/S4.da
..1..2..3.2..1..4.3..5..6....51..2...4..7...12....8.3...96...7..6..4.3..8....7..4 # C27 [asymmetric]
..1..3..2.2..4..3.4..6..5....23...1..4...57..5...9...8..6.8..7..9.1..6..7....2..1 # C27/S2.d
2..1....3..1..2.4..5..3.6..5..7...9...6.8.2...8...4..1..7.6..8..6.4..1..3....7..5 # C27/S4.da
2...3.1....14...2..5...6..3.6.2..4..3....4.7...7.8...65..8..7...9..1..4...6..2..9 # C27/S2.d
6....2.7...8.1...5.9.5..3....2..1..3.1..2..4.5..3..2....4..8.9.1...9.8...3.4....7 # C27.M/S4.da
..1.2...3.2...34..5..1...6...4..1.7.2...9...6.6.4..8...9...2..8..69...1.3...5.7.. # C27.M/S4.da
Serg wrote:Hi, champagne!
Let's refine considered class of pattern:
1. Each pattern contains 18 clues.
2. Each row and each column of the pattern contain 2 clues each exactly.
3. Each minirow and each minicolumn of the pattern contain not more than 1 clue each.
4. You consider essentially different patterns only.
Right?
I would be interesting to see your list of 72 ED patterns of the class defined above. (Do you mean that no other ED patterns of this class do exist?)
Serg