The New Sudoku Players' Forum

[sultan vinegar]

Hi champagne, I have a question on exocets (my favourite pattern). There is a technique in XSUDO called "almost locked candidates". It has a similar flavor to your exocet in that one can argue whatever is true in the base must also be true in the target, but unlike your exocet, there is only 1 cell and two candidates in the base. An example to illustrate:

Code: Select all

361094000098365400040810639637020504000040326024653000050470063006530740473086000

+--------------+-------------+----------------------+
| 3     6   1  | 27   9  4   | (28)    578   2578   |
| 27    9   8  | 3    6  5   | 4       17    127    |
| 257   4   25 | 8    1  27  | 6       3     9      |
+--------------+-------------+----------------------+
| 6     3   7  | 19   2  189 | 5       189   4      |
| 1589  18  59 | 179  4  178 | 3       2     6      |
| 189   2   4  | 6    5  3   | 19-8    1789  178    |
+--------------+-------------+----------------------+
| 1289  5   29 | 4    7  129 | 19(28)  6     3      |
| 1289  18  6  | 5    3  129 | 7       4     -1(28) |
| 4     7   3  | 129  8  6   | 19(2)   159   15     |
+--------------+-------------+----------------------+




Here, the base is r1c7, and the target cell is r8c9. The logic is pretty simple. Whatever is true in the base is eliminated from everywhere in box 9 apart from the target. Hence, as candidate 1 is not a member of the base, it can be eliminated from the target. There is also a secondary elimination of candidate 8 in column 7.

So my question is, would you call this pattern an exocet? If so, it would be a nice introduction to the pattern for beginners without being overwhelmed with so many candidates.

[Leren]

Code: Select all

*--------------------------------------------------------------*
| 3     6     1      | 27    9     4      |a28    578  d2578   |
| 27    9     8      | 3     6     5      | 4     17   d127    |
| 257   4     25     | 8     1     27     | 6     3     9      |
|--------------------+--------------------+--------------------|
| 6     3     7      | 19    2     189    | 5     189   4      |
| 1589  18    59     | 179   4     178    | 3     2     6      |
| 189   2     4      | 6     5     3      | 19-8  1789  178    |
|--------------------+--------------------+--------------------|
| 1289  5     29     | 4     7     129    |b1289  6     3      |
| 1289  18    6      | 5     3     129    | 7     4    c28-1   |
| 4     7     3      | 129   8     6      | 129   159   15     |
*--------------------------------------------------------------*

I'd call this an M Ring Type B (28) - a continuous loop: (2=8) r1c7 - r7c7 = (8-2) r8c9 = r12c9 loop

Loop theory says that for continuous loops, eliminations can be made inside the weak links in the loop (not including the weak link nodes).

These are (1) the 8's in column 7 other than in r7c1 and r7c7 (r6c7 <> 8); (2) The candidates in r8c9 other than 2 and 8 (r2c8 <> 1); and (3) the 2's in box 3 other than the three 2's in the loop (no eliminations).

Hope this helps.

Leren

[ArkieTech]

Code: Select all

 *-----------------------------------------------------------*
 | 3     6     1     | 27    9     4     |a28    578   2578  |
 | 27    9     8     | 3     6     5     | 4     17    127   |
 | 257   4     25    | 8     1     27    | 6     3     9     |
 |-------------------+-------------------+-------------------|
 | 6     3     7     | 19    2     189   | 5     189   4     |
 | 1589  18    59    | 179   4     178   | 3     2     6     |
 | 189   2     4     | 6     5     3     | 19-8  1789  178   |
 |-------------------+-------------------+-------------------|
 | 1289  5     29    | 4     7     129   |b1289  6     3     |
 | 1289  18    6     | 5     3     129   | 7     4     28-1  |
 | 4     7     3     | 129   8     6     |b129  c159  c15    |
 *-----------------------------------------------------------*
Sue de Coq
28r1c7-(28=19)r79c7-519r9c89loop => -1r8c9 -8r6c7

[champagne]

Hi champagne, I have a question on exocets (my favourite pattern). There is a technique in XSUDO called "almost locked candidates". It has a similar flavor to your exocet in that one can argue whatever is true in the base must also be true in the target, but unlike your exocet, there is only 1 cell and two candidates in the base.

So my question is, would you call this pattern an exocet? If so, it would be a nice introduction to the pattern for beginners without being overwhelmed with so many candidates.

The smallest Exocet has a base of 2 digits but 2 targets as in any exocet.

Just keep in mind the basic logic of the exocet

If in it's floor (not considering candidates with another digit) a digit valid in the base forces the target to be filled,
then the target can not contain any other digit that the base digits.

The main reason why that pattern has a big interest is that it is a very common one in the hardest puzzles.
The second reason is that a very common exocet (Jexocet) is relatively easy to detect and verify for a player.

I started an exploration of the area below the hardest (grey zone) to try to figure out whether that pattern remains an interesting solving tool.
The preliminary results let think that the minimum rating to consider is over SE 9.0

So far, the most interesting seen exocets have a base of 3 or 4 digits

[sultan vinegar]

Thanks champagne. So just to confirm, are you saying that the "almost locked candidates" technique does not qualify officially as an (easy example of an) exocet because the minimum requirement for an exocet is two cells in the base and two cells in the target? Note that the "almost locked candiadates" technique does satisfy the logic you wrote in blue text. By the way, I'm not complaining, it's your (brilliant) technique, so if you wish to impose that minimum qualification then that's all good.

Thanks also to ArkieTech and Leren. Of course, this is just a simple continuous AIC (although the Sue de Coq is a nifty interpretation). My motivation was driven by the fact that the JExocet can be described by overlapping remote-finned swordfish, and I was investigating whether (simpler) exocets were possible by using smaller overlapping fish (two overlapping remote-finned Franken cyclops-fish in this case).

[champagne]

By the way, I'm not complaining, it's your (brilliant) technique, so if you wish to impose that minimum qualification then that's all good.

may be some comment on the history of the exocet.

First of all, this is just a translation of an elimination proposed by Allan Barker in the solution of the puzzle "fata morgana". At that time, my solver could not solve that puzzle although all other known puzzles could be solved, including "golden nugget" a very tough one.

So from the beginning, I knew that that property was bringing something not easily shown using chain nets.
The success of that pattern came when it appeared that most of the "hardest puzzles" had it.
Another reason is that each piece of proof is a fish mode analysis, and can never be very difficult.
The focus on the Jexocet made by David P Bird added some interest to the pattern.

The specificity of the logic (something chains don't see easily) requires the 2 targets (with all extensions done later eg: one of the target being an AHS).

In your example, we are not so far from the exocet logic, but classical tools can do the same.

[sultan vinegar]

Yes, I agree that two base cells and two target cells are needed to capture the full essence of the exocet. The "almost locked candidates" is a degenerate case that doesn't really make use of the "higher-order inference" that makes the exocet so special in my view. And, as you say, simpler tools do the job.

I just thought that it would be a good example to use to teach beginners the exocet logic as a first step, because it is so much simpler than say a JExocet (which they could learn as a second step). This is similar to say, first learning about X-Wings before generalizing to Swordfish etc., even though X-Wings can be done with a simple continuous AIC.