The New Sudoku Players' Forum

Website · by **denis_berthier** » Tue Nov 30, 2021 7:35 am

.

Code: Select all: +-------+-------+-------+ ! 1 . . ! . . . ! 7 8 9 ! ! . . . ! . . . ! 2 . . ! ! . . 9 ! . 7 3 ! . . . ! +-------+-------+-------+ ! . . . ! . 3 8 ! 5 . 4 ! ! . . . ! 7 . 5 ! . 9 . ! ! 8 . . ! . 4 . ! . 7 . ! +-------+-------+-------+ ! . 6 . ! . 9 . ! . 4 . ! ! . 3 8 ! . . 4 ! . . . ! ! . . 4 ! . 2 . ! . 5 . ! +-------+-------+-------+ 1.....789......2....9.73.......385.4...7.5.9.8...4..7..6..9..4..38..4.....4.2..5. SER = 8.2

by **Cenoman** » Tue Nov 30, 2021 10:02 am

Code: Select all: +-----------------------+----------------------+------------------------+ | 1 2-5 3 | 4 b56 26 | 7 8 9 | | 4 B57z 567z | 159 8 19 | 2 3 156 | | 256 8 9 | 125 7 3 | 4 16 156 | +-----------------------+----------------------+------------------------+ | 2679 B1279 1267 | 1269 3 8 | 5 126 4 | | 3 4 Aa126y | 7 b16 5 | 168 9 1268 | | 8 B1259 1256 | 1269 4 1269 | 136 7 1236 | +-----------------------+----------------------+------------------------+ | 257 6 1257 | 38 9 17 | 138 4 12378 | | 279 3 8 | 156 156 4 | 169 126 1267 | | 79 B179 4 | 38 2 167 | 13689 5 13678 | +-----------------------+----------------------+------------------------+

1. Death Blossom, stem r5c3
(1)r5c3 - (1=65)r15c5
(2)r5c3 - (2=1795)r2469c2
(6)r5c3 - (6=75)r2c23
=> -5 r1c2; 13 placements & ls

Hidden Text: Show

Code: Select all: +----------------------+-----------------+-------------------+ | 1 2 3 | 4 5 6 | 7 8 9 | | 4 57 567 | 1 8 9 | 2 3 56 | | 56 8 9 | 2 7 3 | 4 16 156 | +----------------------+-----------------+-------------------+ | 2679 179 1267 | 69 3 8 | 5 26 4 | | 3 4 26 | 7 1 5 | 68 9 268 | | 8 59 56 | 69 4 2 | 1 7 3 | +----------------------+-----------------+-------------------+ | 257 6 1257 | 38 9 17 | 38 4 27 | | 27 3 8 | 5 6 4 | 9 12 127 | | 79 179 4 | 38 2 17 | 368 5 68 | +----------------------+-----------------+-------------------+

2. X-Wing (6)r4c1 = r3c1 - r3c8 = r4c8 => -6 r3c9, r4c34

by **marek stefanik** » Tue Nov 30, 2021 10:22 am

My path is not very efficient but uses the same uniqueness-based core in different ways.

Code: Select all: .------------------.-----------------.-------------------. | 1 A25 3 | 4 56 e26 | 7 8 9 | | 4 57 567 | 159 8 19 | 2 3 156 | |c256 8 9 |d125 7 3 | 4 16 156 | :------------------+-----------------+-------------------: | 2679 #1279 1267 |#1269 3 8 | 5 126 4 | | 3 4 126 | 7 16 5 | 168 9 1268 | | 8 #1259 1256 |#1269 4 fα29–16| 136 7 1236 | :------------------+-----------------+-------------------: | 257 6 1257 | 38 9 17 | 138 4 12378 | |b279 3 8 | 156 156 4 | 169 126 1267 | |b79 a179 4 | 38 2 167 | 13689 5 13678 | '------------------'-----------------'-------------------' UR 29c2b5, externals 9r9c2 2r1c2 29r6c6 | 9r9c2 – (9=27)r89c1 – 2r3c1 = 2r3c4 – 2r1c6 = 2r6c6 | 2r1c2 ––––––––––––––––––––––––––––––/e | 29r6c6 => –16r6c6

Code: Select all: .------------------.----------------.-------------------. | 1 25 3 | 4 56 b26 | 7 8 9 | | 4 57 567 | 159 8 d19 | 2 3 156 | | 256 8 9 |a25–1 7 3 | 4 16 156 | :------------------+----------------+-------------------: | 2679 1279 1267 | 1269 3 8 | 5 126 4 | | 3 4 126 | 7 16 5 | 168 9 1268 | | 8 1259 1256 | 1269 4 c29 | 136 7 1236 | :------------------+----------------+-------------------: | 257 6 1257 | 38 9 17 | 138 4 12378 | | 279 3 8 | 156 156 4 | 169 126 1267 | | 79 179 4 | 38 2 167 | 13689 5 13678 | '------------------'----------------'-------------------'

2r3c4 = 2r1c6 – (2=9)r6c6 – (9=1)r2c6 => –1r3c4

Code: Select all: .------------------.----------------.-------------------. | 1 25 3 | 4 56 26 | 7 8 9 | | 4 57 567 |#19 8 #19 | 2 3 56 | |C256 8 9 |D25 7 3 | 4 16 156 | :------------------+----------------+-------------------: | 2679 #1279 1267 |F#1269 3 8 | 5 126 4 | | 3 4 126 | 7 a1–6 5 | 168 9 1268 | | 8 #1259 1256 |F#1269 4 #29 | 136 7 1236 | :------------------+----------------+-------------------: | 257 6 1257 | 38 9 17 | 138 4 12378 | |B279 3 8 |E156 156 4 | 169 126 1267 | |B79 A179 4 | 38 2 167 | 13689 5 13678 | '------------------'----------------'-------------------' DP 19b2 129b5 29c2, externals 1r5c5 9r9c2 2r1c2 (it doesn't matter that 1r6c6 has been eliminated, as long as it was present in the given state) | 1r5c5 | 9r9c2 – (9=27)r89c1 – 2r3c1 = (2–5)r3c4 = (5–6)r8c4 = 6r46c4 | 2r1c2 ––––––––––––––/C => –6r5c5

Code: Select all: .------------------.--------------.-------------------. | 1 a25 3 | 4 56 26 | 7 8 9 | | 4 a57 a567 | 19 8 19 | 2 3 56 | | 256 8 9 | 25 7 3 | 4 16 156 | :------------------+--------------+-------------------: | 2679 179–2 1267 | 269 3 8 | 5 126 4 | | 3 4 b26 | 7 1 5 | 68 9 268 | | 8 159–2 1256 | 269 4 29 | 136 7 1236 | :------------------+--------------+-------------------: | 257 6 1257 | 38 9 17 | 138 4 12378 | | 279 3 8 | 15 56 4 | 169 126 1267 | | 79 179 4 | 38 2 167 | 13689 5 13678 | '------------------'--------------'-------------------'

(2=567)b1p256 – (6=2)r5c3 => –2r46c2

Then solves with an x-wing (6c18\r34).

Marek

by **P.O.** » Tue Nov 30, 2021 10:47 am

Code: Select all: after singles and intersections, two chains in the first state of the grid: 1 25 3 4 56 c2+6 7 8 9 4 57 567 159 8 19 2 3 156 a25+6 8 9 b1+25 7 3 4 16 156 a2±679 1279 12×67 12×69 3 8 5 f12+6 4 3 4 126 7 16 5 168 9 1268 8 1259 1256 1269 4 1269 136 7 1236 257 6 1257 1358 9 d(17) 138 4 12378 2579 3 8 e1(56) e1(56) 4 169 f12-6 1267 79 179 4 1368 2 d(17)6 13689 5 13678 c1n6{r4 r3} - r3n2{c1 c4} - r1c6{n2 n6} - c6{r7r9}{n1n7} - r8{c4c5}{n5n6} - c8n6{r8 r4} => r4c3 r4c4 <> 6 1 a±2×5 3 4 e+56 26 7 8 9 4 57 567 159 8 19 2 3 156 a+256 8 9 125 7 3 4 16 156 b2+679 a1*279 1267 1269 3 8 5 126 4 3 4 c+126 7 d1+6 5 168 9 1268 8 a1*259 1256 1269 4 1269 136 7 1236 257 6 1257 1358 9 17 138 4 12378 2579 3 8 156 156 4 169 126 1267 79 179 4 1368 2 167 13689 5 13678 c2n2{r1 r4r6} -\ 》c1n6{r3 r4} - r5c3{n2n6 n1} - r5c5{n1 n6} - r1c5{n6 n5} => r1c2 <> 5 b1n2{r1c2 r3c1} -/ ste.

Hi Denis, i use regularly your collection of puzzles to test my algorithm as you gave with them their SER and W rating; solving this puzzle and seeing the first row is the same as all (?) of yours i search and found it as the 7191th; the point is this puzzle is a example of the discrepancy between rating systems: as a SER 8.2 it is in the difficult category but as a W-rating of 3 it is rather a easy one, my simplest-first path is 11 chains with max length 3.

by **AnotherLife** » Tue Nov 30, 2021 12:14 pm

P.O. wrote:... the point is this puzzle is a example of the discrepancy between rating systems: as a SER 8.2 it is in the difficult category but as a W-rating of 3 it is rather a easy one...

Hello, P.O.
I agree with you. This is also a discrepancy between ER and my rating system. If you read my recent posts, you can find many examples of inflated ratings by the ER system. I rate the current puzzle as class GC because it is solvable via AICs with groups. Here is a simple solution.

Code: Select all: .------------------.------------------.-------------------. | 1 25 3 | 4 a56 26 | 7 8 9 | | 4 57 567 | 159 8 19 | 2 3 156 | | 256 8 9 | 125 7 3 | 4 16 156 | :------------------+------------------+-------------------: | 2679 1279 1267 | d1269 3 8 | 5 126 4 | | 3 4 126 | 7 1-6 5 | 168 9 1268 | | 8 1259 1256 | d1269 4 1269 | 136 7 1236 | :------------------+------------------+-------------------: | 257 6 1257 | 38 9 17 | 138 4 12378 | | 279 3 8 | c156 b156 4 | 169 126 1267 | | 79 179 4 | 38 2 167 | 13689 5 13678 | '------------------'------------------'-------------------'

1. AIC with a group
(6=5)r1c5 - r8c5 = (5-6)r8c4 = r46c4 => -6 r5c5; 1 single

Code: Select all: .--------------------.--------------.-------------------. | 1 a25 3 | 4 56 26 | 7 8 9 | | 4 57 567 | 159 8 19 | 2 3 156 | | b256 8 9 | 125 7 3 | 4 16 156 | :--------------------+--------------+-------------------: | c2679 179-2 1267 | 269 3 8 | 5 126 4 | | 3 4 d26 | 7 1 5 | 68 9 268 | | 8 159-2 1256 | 269 4 269 | 136 7 1236 | :--------------------+--------------+-------------------: | 257 6 1257 | 38 9 17 | 138 4 12378 | | 279 3 8 | 156 56 4 | 169 126 1267 | | 79 179 4 | 38 2 167 | 13689 5 13678 | '--------------------'--------------'-------------------'

2. Basic AIC
(2)r1c2 = (2-6)r3c1 = r4c1 - (6=2)r5c3 => -2 r46c2; 12 singles and lcls steps

Code: Select all: .------------------.-----------.----------------. | 1 2 3 | 4 5 6 | 7 8 9 | | 4 57 567 | 1 8 9 | 2 3 56 | | 56* 8 9 | 2 7 3 | 4 16* 15-6 | :------------------+-----------+----------------: | 2679* 179 127-6 | 9-6 3 8 | 5 26* 4 | | 3 4 26 | 7 1 5 | 68 9 268 | | 8 59 56 | 69 4 2 | 1 7 3 | :------------------+-----------+----------------: | 257 6 1257 | 38 9 17 | 38 4 27 | | 27 3 8 | 5 6 4 | 9 12 127 | | 79 179 4 | 38 2 17 | 368 5 68 | '------------------'-----------'----------------'

3. X-Wing (6)c18 r34 => -6 r3c9, r4c34; ste

Website · by **denis_berthier** » Tue Nov 30, 2021 12:32 pm

P.O. wrote: i use regularly your collection of puzzles to test my algorithm as you gave with them their SER and W rating; solving this puzzle and seeing the first row is the same as all (?) of yours i search and found it as the 7191th; the point is this puzzle is a example of the discrepancy between rating systems: as a SER 8.2 it is in the difficult category but as a W-rating of 3 it is rather a easy one, my simplest-first path is 11 chains with max length 3.

Hi P.O.
The discrepancy is still higher, as the puzzle is not only in W3 but in Z3.

Code: Select all: Resolution state after Singles and whips[1]: +-------------------+-------------------+-------------------+ ! 1 25 3 ! 4 56 26 ! 7 8 9 ! ! 4 57 567 ! 159 8 19 ! 2 3 156 ! ! 256 8 9 ! 125 7 3 ! 4 16 156 ! +-------------------+-------------------+-------------------+ ! 2679 1279 1267 ! 1269 3 8 ! 5 126 4 ! ! 3 4 126 ! 7 16 5 ! 168 9 1268 ! ! 8 1259 1256 ! 1269 4 1269 ! 136 7 1236 ! +-------------------+-------------------+-------------------+ ! 257 6 1257 ! 1358 9 17 ! 138 4 12378 ! ! 2579 3 8 ! 156 156 4 ! 169 126 1267 ! ! 79 179 4 ! 1368 2 167 ! 13689 5 13678 ! +-------------------+-------------------+-------------------+ 147 candidates.

Code: Select all: hidden-pairs-in-a-column: c4{n3 n8}{r7 r9} ==> r9c4≠6, r9c4≠1, r7c4≠5, r7c4≠1 whip[1]: b8n5{r8c5 .} ==> r8c1≠5 finned-x-wing-in-columns: n2{c6 c2}{r1 r6} ==> r6c3≠2 biv-chain[3]: r5c5{n1 n6} - b2n6{r1c5 r1c6} - c6n2{r1 r6} ==> r6c6≠1 z-chain[3]: c5n1{r5 r8} - r8n5{c5 c4} - c4n6{r8 .} ==> r5c5≠6 naked-single ==> r5c5=1 finned-x-wing-in-rows: n6{r2 r5}{c3 c9} ==> r6c9≠6 biv-chain[3]: r5c3{n2 n6} - c1n6{r4 r3} - b1n2{r3c1 r1c2} ==> r4c2≠2, r6c2≠2 singles ==> r1c2=2, r1c6=6, r1c5=5, r8c5=6, r8c4=5, r3c4=2, r2c4=1, r2c6=9, r6c6=2 whip[1]: r8n1{c9 .} ==> r7c7≠1, r7c9≠1, r9c7≠1, r9c9≠1 naked-pairs-in-a-row: r7{c4 c7}{n3 n8} ==> r7c9≠8, r7c9≠3 x-wing-in-columns: n6{c1 c8}{r3 r4} ==> r4c4≠6, r4c3≠6, r3c9≠6 stte

Moreover, it has several 2-step solutions in Z6, e.g.

Code: Select all: z-chain[6]: b1n2{r1c2 r3c1} - b1n6{r3c1 r2c3} - r5c3{n6 n1} - c5n1{r5 r8} - c5n5{r8 r1} - r1c2{n5 .} ==> r4c2≠2, r6c2≠2 with z-candidates = n2r5c3 n2r1c2 singles ==> r1c2=2, r1c6=6, r1c5=5, r3c4=2, r6c6=2, r2c6=9, r2c4=1, r5c5=1, r8c5=6, r8c4=5 whip[1]: r8n1{c9 .} ==> r7c7≠1, r7c9≠1, r9c7≠1, r9c9≠1 biv-chain[4]: r3c1{n6 n5} - c2n5{r2 r6} - r6n9{c2 c4} - b5n6{r6c4 r4c4} ==> r4c1≠6 stte

A rating is good as long as it is used to rate puzzles you intend to solve with the rules it takes into account. This is an example of a tautology.
The SER is a good rating, in the mean, even if you use different rules. Of, course, one will find exceptions.
People keep talking about ALS chains, but I still have to see a rating based on them and statistical results (e.g. for the cbg-000 collection).

by **AnotherLife** » Tue Nov 30, 2021 1:04 pm

denis_berthier wrote:[
People keep talking about ALS chains, but I still have to see a rating based on them and statistical results (e.g. for the cbg-000 collection).

Hello, Denis,
Where can I find this cbg-000 collection?

Website · by **denis_berthier** » Tue Nov 30, 2021 1:45 pm

AnotherLife wrote:
denis_berthier wrote:[People keep talking about ALS chains, but I still have to see a rating based on them and statistical results (e.g. for the cbg-000 collection).

Where can I find this cbg-000 collection?

in the Sudoku examples folder of CSP-Rules-V2.1, together with many ratings for it: https://github.com/denis-berthier/CSP-Rules-V2.1/tree/master/Examples/Sudoku/cbg-000
You can also find the full controlled-bias collection of ~6M puzzles and their W ratings here: https://github.com/denis-berthier/Controlled-bias_Sudoku_generator_and_collection But it has only about 1M of the SER ratings.

by **AnotherLife** » Tue Nov 30, 2021 5:31 pm

My preliminary analysis of the cbg-000 collection confirms my assumption that in many cases ER ratings are inflated. Among the first 413 puzzles of the list, only one puzzle with ER<=8.3

Code: Select all: .2.4..7......891.........65..48.....3..9....1.95..1.7..7.3...1263.........2.1...8

really needs a forcing chain to be solved, the other puzzles being solved via standard patterns, AICs with or without groups and almost locked sets. As I cannot run the checking process automatically, I am unable to analyze the entire list of 3,037,717 puzzles, but I will try to get a longer list of puzzles of the ALSC-class.

Website · by **denis_berthier** » Tue Nov 30, 2021 7:12 pm

AnotherLife wrote:My preliminary analysis of the cbg-000 collection confirms my assumption that in many cases ER ratings are inflated.

Why am I sure any analysis you will make will confirm your prejudices?
How can the SER be inflated? The SER is what it is.
You want to promote another rating? OK, but then you'd better start by defining it. This will require more than a series of vague claims along with the solution of a few puzzles.

AnotherLife wrote:Among the first 413 puzzles of the list, only one puzzle with ER<=8.3 really needs a forcing chain to be solved, the other puzzles being solved via standard patterns, AICs with or without groups and almost locked sets.

How can you reach this result manually in less than 4 hrs? That makes 100 puzzles per hour, solved manually with patterns that are not in SE.

AnotherLife wrote:puzzles of the ALSC-class.

Nobody knows what this is.

by **AnotherLife** » Tue Nov 30, 2021 8:18 pm

denis_berthier wrote:How can the SER be inflated?

In many cases the SER claims that some hard methods are required to solve the puzzle, but in fact the puzzle is solvable by simpler methods.

denis_berthier wrote:You want to promote another rating? OK, but then you'd better start by defining it.

I don't know where I can write about it. Will you give me advice?

denis_berthier wrote:
AnotherLife wrote:Among the first 413 puzzles of the list, only one puzzle with ER<=8.3 really needs a forcing chain to be solved, the other puzzles being solved via standard patterns, AICs with or without groups and almost locked sets.

How can you reach this result manually in less than 4 hrs? That makes 100 puzzles per hour, solved manually with patterns that are not in SE.

I have to check only the puzzles rated 7.9<=ER<=9.0 (27 puzzles), and the simpler puzzles are solvable by standard patterns and basic AIC's.

denis_berthier wrote:
AnotherLife wrote:puzzles of the ALSC-class.

Nobody knows what this is.

By ALSC-class I mean all the puzzles solvable by standard patterns, basic AICs, AICs with groups and ALS's.

I have started to analyze a longer list of more than 10, 000 puzzles, and later I will tell you about my results. Also I will post some new 'fake monsters'.

Website · by **denis_berthier** » Wed Dec 01, 2021 4:28 am

AnotherLife wrote:
denis_berthier wrote:How can the SER be inflated?

In many cases the SER claims that some hard methods are required to solve the puzzle, but in fact the puzzle is solvable by simpler methods.

This is not what the SER claims. It claims that you get this rating IF you use only these methods (like any rating).

AnotherLife wrote:
denis_berthier wrote:You want to promote another rating? OK, but then you'd better start by defining it.
I don't know where I can write about it. Will you give me advice?

If you have something to say about a new rating system, a thread in the GENERAL section would IMO be the best place.

AnotherLife wrote:
denis_berthier wrote:
AnotherLife wrote:Among the first 413 puzzles of the list, only one puzzle with ER<=8.3 really needs a forcing chain to be solved, the other puzzles being solved via standard patterns, AICs with or without groups and almost locked sets.
How can you reach this result manually in less than 4 hrs? That makes 100 puzzles per hour, solved manually with patterns that are not in SE.

I have to check only the puzzles rated 7.9<=ER<=9.0 (27 puzzles), and the simpler puzzles are solvable by standard patterns and basic AIC's.

I think you need to check many more puzzles, unless you want your rating to be a hotpot of unrelated techniques.

Code: Select all: 4.5 UR Types 1 or 2 or 4 or 3 w/ hidden pair 4.6 UR Type 3 w/ naked pair or hidden triplet UL Types 1 or 2 or 4 or 3 w/ hidden pair (6 cells) 4.69 UL Type 3 w/ a naked pair or hidden triplet (6 cells) 4.7 UR Type 3 w/ naked triplet or hidden quad UL Types 1 or 2 or 4 or 3 w/ hidden pair (8 cells) 4.8 UR Type 3 w/ naked quad UL Type 3 w/ naked triplet [or hidden quad] (6 cells) UL Type 3 w/ naked pair or hidden triplet (8 cells) 4.89 UL Type 3 w/ naked quad (6 cells) 4.9 [UL Type 3 w/ naked triplet or hidden quad (8 cells)] 5.0 Naked Quad or UL 1 or 2 or 4 (>=10 cells) 5.1 UL Type 3 w/ naked pair (>=10 cells) 5.6 BUG Type 1 5.7 BUG Type 2 or 4 5.8 BUG Type 3 w/ naked pair 5.9 BUG Type 3 w/ naked triplet 6.0 BUG Type 3 w/ naked quad 6.1 BUG Type 3 w/ naked quint 6.2 Aligned Pair Exclusion 6.6 Turbot Fish Forcing X-chain or Bidirectional Y-Cycle (5-6 nodes) 6.69 Forcing X-Chain (7-8 nodes) 6.7 Bidirectional Y-cycle (7-8 nodes) 6.8 Forcing X-Chain or Bidirectional Y-cycle (9-12 nodes) 6.9 Forcing X-Chain or Bidirectional Y-cycle (13-16 nodes) 7.0 Bidirectional Y-cycle (17-24 nodes) Forcing Chain or Bidirectional Cycle (1-4 nodes) 7.1 Forcing Chain or Bidirectional Cycle (5-6 nodes) 7.2 Forcing Chain or Bidirectional Cycle (7-8 nodes) 7.3 Forcing Chain or Bidirectional Cycle (9-12 nodes) 7.4 Forcing Chain (13-16 nodes) 7.5 Forcing Chain (17-24 nodes) Aligned Triplet Exclusion 7.6 Forcing Chain (25-36 nodes) Nishio Forcing Chain (5-6 nodes) 7.7 Nishio Forcing Chain (7-8 nodes) 7.8 Nishio Forcing Chain (9-12 nodes) 7.9 Nishio Forcing Chain (13-16 nodes)

The New Sudoku Players' Forum

Puzzle-8.2

Puzzle-8.2

Re: Puzzle-8.2

Re: Puzzle-8.2

Re: Puzzle-8.2

Re: Puzzle-8.2

Re: Puzzle-8.2

Re: Puzzle-8.2

Re: Puzzle-8.2

Re: Puzzle-8.2

Re: Puzzle-8.2

Re: Puzzle-8.2

Re: Puzzle-8.2