This guide will have four sections: Basics, Triplets, Virtual Sets, Cannibalism.
Basics:
Hidden Text: Show
BASICS
Truth-Set: A collection of elements, at least one of which must be true. Truth-sets are typically one of the four native Sudoku sets; row (R), column (C), box (B), cell (N). In rare cases, one can also use virtual (V) sets, which are described in a separate section.
Link-set: A collection of elements, at most one of which must be true. Link-sets are always one of the four native Sudoku types; R,C,B,N. In theory, you could use a virtual link-set, but I have never seen it done, and I believe that xsudo does not have the facility to do it.
Logic: The collection of all truth-sets and link-sets. The only caveat is that all elements of all truth-sets must also be elements of at least one link-set. This is in contrast with the definition of cover-set in the Ultimate Fish Guide (UFG), but consistent with the definition of cover set in Hobiwan fish.
Triplet: There are two types of triplet:
A truth-set-triplet is an element that is a member of two truth-sets and one link-set.
A link-set triplet is an element that is a member of one truth-set and two link-sets.
Triplets will be described in a separate section. The basic techniques assume that there are no triplets in the logic.
Rank: In the absence of any triplets:
Rank is a number defined as the number of link-sets in the logic minus the number of truth-sets in the logic.
There is a connection between rank and the number of overlapping link-sets that guarantee the elimination of a candidate.
Theorem: For rank R logic, any candidate that is a member of 0 truth-sets and R+1 or more link-sets is guaranteed to be false, and maybe eliminated.
Proof:
For definiteness, let the number of truth-sets be “n”. There are no truth-set triplets by assumption, so there are thus at least “n” true candidates in the collection of truth-sets.
As the rank is R, the number of link-sets is “n+R”.
Assume that a candidate “x” that is a member of 0 truth-sets and R+1 or more link-sets is true.
Let’s tally up how many of our “n+R” link-sets are guaranteed to contain a true candidate. As each member of each truth-set must also be a member of exactly one link-set (no triplets remember), at least “n” link-sets contain the at least “n” true candidates from the collection of “n” truth-sets (remember, each link-set can contain at most one true candidate). In addition, “R+1” link-sets contain true candidate “x” by assumption. These “R+1” link-sets are separate to the “n” already mentioned, as there are no triplets, so no opportunity for overlap. In total, we have “n+R+1” link-sets containing a true candidate.
But we have only “n+R” link-sets, so we have a contradiction, and the assumption of true candidate “x” must be false. Hence, “x” can be eliminated.
As there are only 4 types of native link-set, there is a maximum of 4 overlapping link-sets, so the maximum rank that guarantees an elimination is rank 3. For ranks higher than 3, there will need to be at least one triplet present if an elimination is to occur.
Examples:
Rank 0:
Rank 1:
Rank 2:
Rank 3:
Truth-Set: A collection of elements, at least one of which must be true. Truth-sets are typically one of the four native Sudoku sets; row (R), column (C), box (B), cell (N). In rare cases, one can also use virtual (V) sets, which are described in a separate section.
Link-set: A collection of elements, at most one of which must be true. Link-sets are always one of the four native Sudoku types; R,C,B,N. In theory, you could use a virtual link-set, but I have never seen it done, and I believe that xsudo does not have the facility to do it.
Logic: The collection of all truth-sets and link-sets. The only caveat is that all elements of all truth-sets must also be elements of at least one link-set. This is in contrast with the definition of cover-set in the Ultimate Fish Guide (UFG), but consistent with the definition of cover set in Hobiwan fish.
Triplet: There are two types of triplet:
A truth-set-triplet is an element that is a member of two truth-sets and one link-set.
A link-set triplet is an element that is a member of one truth-set and two link-sets.
Triplets will be described in a separate section. The basic techniques assume that there are no triplets in the logic.
Rank: In the absence of any triplets:
Rank is a number defined as the number of link-sets in the logic minus the number of truth-sets in the logic.
There is a connection between rank and the number of overlapping link-sets that guarantee the elimination of a candidate.
Theorem: For rank R logic, any candidate that is a member of 0 truth-sets and R+1 or more link-sets is guaranteed to be false, and maybe eliminated.
Proof:
For definiteness, let the number of truth-sets be “n”. There are no truth-set triplets by assumption, so there are thus at least “n” true candidates in the collection of truth-sets.
As the rank is R, the number of link-sets is “n+R”.
Assume that a candidate “x” that is a member of 0 truth-sets and R+1 or more link-sets is true.
Let’s tally up how many of our “n+R” link-sets are guaranteed to contain a true candidate. As each member of each truth-set must also be a member of exactly one link-set (no triplets remember), at least “n” link-sets contain the at least “n” true candidates from the collection of “n” truth-sets (remember, each link-set can contain at most one true candidate). In addition, “R+1” link-sets contain true candidate “x” by assumption. These “R+1” link-sets are separate to the “n” already mentioned, as there are no triplets, so no opportunity for overlap. In total, we have “n+R+1” link-sets containing a true candidate.
But we have only “n+R” link-sets, so we have a contradiction, and the assumption of true candidate “x” must be false. Hence, “x” can be eliminated.
As there are only 4 types of native link-set, there is a maximum of 4 overlapping link-sets, so the maximum rank that guarantees an elimination is rank 3. For ranks higher than 3, there will need to be at least one triplet present if an elimination is to occur.
Examples:
Rank 0:
Rank 1:
Rank 2:
Rank 3:
Triplets:
Hidden Text: Show
TRIPLETS
Truth-set Triplet (rare): 2 overlapping truth-sets and 1 overlapping link-set
Link-set Triplet (more common): 1 overlapping truth-set and 2 overlapping link-sets
Triplet State: Occupied (candidate is assumed true) / Unoccupied (candidate is assumed false)
If triplets are present, then the definition of rank is adjusted so that the connection between rank and the number of overlapping link-sets that guarantee the elimination of a candidate is maintained. In general, the rank is not constant as in the case of no triplets; the rank will be different in different regions if triplets are present.
The procedure to calculate the rank is to consider in turn each combination of triplet states. As each triplet has two possible states, for “n” triplets there are 2^n possible combinations of triplet states. For example, if there are six triplets as in the case of the exocet elimination in Fata Morgana, then there are 2^6 or 64 combinations to consider. This is a laborious calculation in general, but is often simplified by “common sense”.
This current section is split into six case studies, showing how to handle triplets, culminating in two “no-fish” examples.
1) Truth-set triplet easy:
In this case, the truth-set triplet behaves identical to an endo-fin from the UFG.
We have 5 truth-sets, 5 link-sets and one truth-set triplet at row 3, column 3, number 2. As there is just one triplet, we have just two possibilities; triplet occupied, triplet unoccupied.
If the triplet is unoccupied, then the logic contains 5 truth-sets, 5 link-sets and no triplet (it has been set to unoccupied). So we have rank 0 logic (the fish is true).
If the triplet is occupied, then the triplet’s link-set is occupied, and so the triplet’s link-set is defined as rank 0 to maintain the connection between rank and the number of overlapping link-sets that guarantee the elimination of a candidate.
So, as in the case of an endo-fin, regardless of the triplet’s state, the triplet’s link-set is rank 0, and any candidate that is a member of 0 sets and the triplet’s link-set can be eliminated.
2) Truth-set triplet hard
In this case, the truth-set triplet does more than what an endo-fin can.
We have 4 truth-sets, 4 link-sets and one truth-set triplet at row 4, column 5, number 1. As there is just one triplet, we have just two possibilities; triplet occupied, triplet unoccupied.
If we use the endo-fin approach, then we discover that the link-set in row 4 is rank 0. But the truth-set triplet can do more than an endo-fin can. I now show that the link-set in row 2 is rank 0, although it doesn’t result in any extra eliminations unfortunately. To do this, we reduce the logic. (Note that I omitted this step in the previous example, as it didn’t show anything new in that case, but here it does.)
Reducing the logic occurs in the case where the truth-set triplet is assumed occupied. We now re-draw the logic with the truth-set triplet’s two sets and one link-set removed (as they have been accounted for in the assumption that the triplet is occupied). We don’t rearrange or add in any new truth-sets or link-sets, we just look at what remains upon deletion of the triplets truth-sets and link-set.
Now, when the triplet was assumed to be unoccupied, the logic was calculated to be rank 0 everywhere. So, when the triplet is assumed to be occupied, we want to match this rank for useful eliminations. The question now is: in this reduced logic, is there a sub-collection of truth-sets and link-sets (satisfying the property that every member of a truth-set is also a member of a link-set) that results in a rank 0 region? Yes! We take set 1C7, and link-set 1R2. Thus, if the triplet is occupied, then the link-set in row 2 is rank 0. Notice that the remaining two link-sets in rows 6 and 8 are rank 1, since that region has one truth-set and two link-sets.
So, regardless of the state of the triplet, the link-sets in rows 2 and 4 are rank 0.
We now turn to link-set triplets. These are similar to exo-fins in the UFG, only link-set triplets can do more than exo-fins can.
3) Link-set triplet easy
We have 5 truth-sets, 7 link-sets and 1 link-set triplet in row 1, column 6, number 7. As there is just one triplet, we have just two possibilities; triplet occupied, triplet unoccupied.
If the triplet is assumed occupied, then we reduce the logic by removing the triplet's truth-set and two link-sets from the diagram. This leaves 4 truth-sets, 5 link-sets, no triplets, so rank 1 logic. As the candidate in row 9, column 9, number 7 is a member of two link-sets, it is eliminated if the triplet is occupied.
If the triplet is assumed unoccupied, then we reduce the logic by removing the candidate in row 1, column 6, number 7 from the diagram since it has been assumed false. Then, as in the case of the truth-set triplet hard, we look for a sub-collection of truth-sets and link-sets (satisfying the property that every member of a truth-set is also a member of a link-set) that results in a rank 1 or lower region (since we want to at least match the rank 1 obtained from the case where the triplet was assumed occupied).
We have 3 truth-sets; 1N6, 1B3, 5C9, and 3 link-sets; 1R1, 3N9, 9N9 that form a rank 0 sub-region. So, all up: if the triplet is assumed occupied, then link-set 9N9 is rank 1. If the triplet is unoccupied, then link-set 9N9 is rank 0. The worst case is rank 1, and candidate 7 in row 9, column 9 is eliminated by (reduced) rank 1 logic.
As a rule, this case is easy as the triplet’s truth-set was a bivalue (a bilocation would be just as easy), and so when the triplet is assumed false, the other element in the triplet’s truth-set must be true. In addition, we get a nice simple chain of inferences to show that if the triplet is unoccupied, then row 9, column 9, number 7 is false. So, regardless of the state of the triplet, row 9, column 9, number 7 is false.
4) Link-set triplet hard
In this case, we apply the same procedure as for the previous link-set triplet case, but this current case is considered hard because we do not get the bivalue/bilocation and chain short-cut that occurred in the previous case.
We have 4 truth-sets, 6 link-sets and a link-set triplet at row 5, column 4, number 6. If the triplet is occupied, then we remove one truth-set and two link-sets from the logic, leaving 3 truth-sets, 4 link-sets and no triplets, so all link-sets are rank 1 if the triplet is occupied. The candidate at row 6, column 9, number 6 is thus eliminated if the triplet is occupied, by rank 1 logic. If the triplet is unoccupied, we remove the candidate at row 5, column 5, number 6 from the logic. We then look at the remaining logic to see if any rank 1 or lower regions exist (so as to at least match the rank 1 elimination from the case where the triplet was occupied). The logic now looks like:
The sub-collection consisting of two truth-sets (6R2, 6B5) and three link-sets (6C5, 6C9, 6R6) is a rank 1 sub-region. So, again, the candidate at row 6, column 9, number 6 is eliminated by rank 1 logic. All up, regardless of the state of the triplet, the candidate at row 6, column 9, number 6 is eliminated by rank 1 logic.
The final two cases have multiple triplets. These two cases are currently under discussion in the no-fish thread, as they are template eliminations with no (classical) fish explanation. These eliminations are explainable with Xsudo logic however.
5) No-fish #1
We have 6 truth-sets, 8 link-sets, 1 truth-set triplet (R9C1) and 2 link-set triplets (R8C3, R9C3). As there are 3 triplets, there are 2^3 or 8 possible triplet states. However, with common sense, we won’t have to check all 8 states, as all triplets share a box, so at most one of them is true.
Case I: The truth-set triplet is occupied, and the two link-set triplets are unoccupied.
This case is easy. Since the truth-set triplet is occupied, the truth-set triplet’s link-set is occupied, hence the link-set in row 9 is rank 0.
Case II: The link-set triplet at R8C3 is occupied, and the two others are unoccupied.
This case is also straight-forward. The occupied triplet’s two link-sets are rank 0. There are 5 truth-sets and 6 link-sets remaining. So each of the remaining link-sets is rank 1. Of interest to us is that the link-set in row 9 is rank 1.
Case III: The link-set triplet at R9C3 is occupied, and the two others are unoccupied.
This case is similar to case II. The occupied triplet’s two link-sets are rank 0. There are 5 truth-sets and 6 link-sets remaining. So each of the remaining link-sets is rank 1. Of interest to us is that the link-set in row 9 is rank 0.
Case IV: All triplets are unoccupied.
In this final case, we reduce the logic following the usual procedure. The logic is redrawn with all triplets unoccupied:
We look for a rank 1 or lower sub-region to at least match our worst case so far which is rank 1. Consider the sub-region with one truth-set (9B7) and one link-set (9R9). This is a rank 0 sub-region, which is good enough for us today.
All up, regardless of the triplets’ state, the worst case rank for the link-set in row 9 is rank 1, so the candidate at row 9, column 8, number 9 is eliminated by (reduced) rank 1 logic.
6) No-fish #2
We have 5 truth-sets, 7 link-sets, 1 truth-set triplet (R2C6) and 2 link-set triplets (R2C1, R2C5). As there are 3 triplets, there are 2^3 or 8 possible triplet states. However, with common sense, we won’t have to check all 8 states, as all triplets share a row, so at most one of them is true.
Case I: The link-set triplet at R2C1 is occupied, and the other two triplets are unoccupied.
The occupied triplet’s two link-sets are rank 0. There are 4 truth-sets and 5 link-sets remaining. So each of the remaining link-sets is rank 1. Of interest to us is that the link-set in row 9 is rank 1.
Case II: The link-set triplet at R2C5 is occupied, and the other two triplets are unoccupied.
The occupied triplet’s two link-sets are rank 0. There are 4 truth-sets and 5 link-sets remaining. So each of the remaining link-sets is rank 1. Of interest to us is that the link-set in row 9 is rank 1.
Case III: The truth-set triplet is occupied, and the other two triplets are unoccupied.
We re-draw the logic after setting the triplet’s to the assumed states.
We look for a sub-region that is rank 1 or lower, to at least match the worst case so far which is rank 1. The region with one truth-set (9C1) and 1 link-set (9R9) is rank 0.
Case IV: All triplets are unoccupied:
We re-draw the logic after setting the triplet’s to the assumed states.
We look for a sub-region that is rank 1 or lower, to at least match the worst case so far which is rank 1. The region with one set (9C1) and 1 link-set (9R9) is rank 0.
All up, regardless of the triplets’ state, the worst case rank for the link-set in row 9 is rank 1, so the candidate at row 9, column 5, number 7 is eliminated by (reduced) rank 1 logic.
Truth-set Triplet (rare): 2 overlapping truth-sets and 1 overlapping link-set
Link-set Triplet (more common): 1 overlapping truth-set and 2 overlapping link-sets
Triplet State: Occupied (candidate is assumed true) / Unoccupied (candidate is assumed false)
If triplets are present, then the definition of rank is adjusted so that the connection between rank and the number of overlapping link-sets that guarantee the elimination of a candidate is maintained. In general, the rank is not constant as in the case of no triplets; the rank will be different in different regions if triplets are present.
The procedure to calculate the rank is to consider in turn each combination of triplet states. As each triplet has two possible states, for “n” triplets there are 2^n possible combinations of triplet states. For example, if there are six triplets as in the case of the exocet elimination in Fata Morgana, then there are 2^6 or 64 combinations to consider. This is a laborious calculation in general, but is often simplified by “common sense”.
This current section is split into six case studies, showing how to handle triplets, culminating in two “no-fish” examples.
1) Truth-set triplet easy:
In this case, the truth-set triplet behaves identical to an endo-fin from the UFG.
We have 5 truth-sets, 5 link-sets and one truth-set triplet at row 3, column 3, number 2. As there is just one triplet, we have just two possibilities; triplet occupied, triplet unoccupied.
If the triplet is unoccupied, then the logic contains 5 truth-sets, 5 link-sets and no triplet (it has been set to unoccupied). So we have rank 0 logic (the fish is true).
If the triplet is occupied, then the triplet’s link-set is occupied, and so the triplet’s link-set is defined as rank 0 to maintain the connection between rank and the number of overlapping link-sets that guarantee the elimination of a candidate.
So, as in the case of an endo-fin, regardless of the triplet’s state, the triplet’s link-set is rank 0, and any candidate that is a member of 0 sets and the triplet’s link-set can be eliminated.
2) Truth-set triplet hard
In this case, the truth-set triplet does more than what an endo-fin can.
We have 4 truth-sets, 4 link-sets and one truth-set triplet at row 4, column 5, number 1. As there is just one triplet, we have just two possibilities; triplet occupied, triplet unoccupied.
If we use the endo-fin approach, then we discover that the link-set in row 4 is rank 0. But the truth-set triplet can do more than an endo-fin can. I now show that the link-set in row 2 is rank 0, although it doesn’t result in any extra eliminations unfortunately. To do this, we reduce the logic. (Note that I omitted this step in the previous example, as it didn’t show anything new in that case, but here it does.)
Reducing the logic occurs in the case where the truth-set triplet is assumed occupied. We now re-draw the logic with the truth-set triplet’s two sets and one link-set removed (as they have been accounted for in the assumption that the triplet is occupied). We don’t rearrange or add in any new truth-sets or link-sets, we just look at what remains upon deletion of the triplets truth-sets and link-set.
Now, when the triplet was assumed to be unoccupied, the logic was calculated to be rank 0 everywhere. So, when the triplet is assumed to be occupied, we want to match this rank for useful eliminations. The question now is: in this reduced logic, is there a sub-collection of truth-sets and link-sets (satisfying the property that every member of a truth-set is also a member of a link-set) that results in a rank 0 region? Yes! We take set 1C7, and link-set 1R2. Thus, if the triplet is occupied, then the link-set in row 2 is rank 0. Notice that the remaining two link-sets in rows 6 and 8 are rank 1, since that region has one truth-set and two link-sets.
So, regardless of the state of the triplet, the link-sets in rows 2 and 4 are rank 0.
We now turn to link-set triplets. These are similar to exo-fins in the UFG, only link-set triplets can do more than exo-fins can.
3) Link-set triplet easy
We have 5 truth-sets, 7 link-sets and 1 link-set triplet in row 1, column 6, number 7. As there is just one triplet, we have just two possibilities; triplet occupied, triplet unoccupied.
If the triplet is assumed occupied, then we reduce the logic by removing the triplet's truth-set and two link-sets from the diagram. This leaves 4 truth-sets, 5 link-sets, no triplets, so rank 1 logic. As the candidate in row 9, column 9, number 7 is a member of two link-sets, it is eliminated if the triplet is occupied.
If the triplet is assumed unoccupied, then we reduce the logic by removing the candidate in row 1, column 6, number 7 from the diagram since it has been assumed false. Then, as in the case of the truth-set triplet hard, we look for a sub-collection of truth-sets and link-sets (satisfying the property that every member of a truth-set is also a member of a link-set) that results in a rank 1 or lower region (since we want to at least match the rank 1 obtained from the case where the triplet was assumed occupied).
We have 3 truth-sets; 1N6, 1B3, 5C9, and 3 link-sets; 1R1, 3N9, 9N9 that form a rank 0 sub-region. So, all up: if the triplet is assumed occupied, then link-set 9N9 is rank 1. If the triplet is unoccupied, then link-set 9N9 is rank 0. The worst case is rank 1, and candidate 7 in row 9, column 9 is eliminated by (reduced) rank 1 logic.
As a rule, this case is easy as the triplet’s truth-set was a bivalue (a bilocation would be just as easy), and so when the triplet is assumed false, the other element in the triplet’s truth-set must be true. In addition, we get a nice simple chain of inferences to show that if the triplet is unoccupied, then row 9, column 9, number 7 is false. So, regardless of the state of the triplet, row 9, column 9, number 7 is false.
4) Link-set triplet hard
In this case, we apply the same procedure as for the previous link-set triplet case, but this current case is considered hard because we do not get the bivalue/bilocation and chain short-cut that occurred in the previous case.
We have 4 truth-sets, 6 link-sets and a link-set triplet at row 5, column 4, number 6. If the triplet is occupied, then we remove one truth-set and two link-sets from the logic, leaving 3 truth-sets, 4 link-sets and no triplets, so all link-sets are rank 1 if the triplet is occupied. The candidate at row 6, column 9, number 6 is thus eliminated if the triplet is occupied, by rank 1 logic. If the triplet is unoccupied, we remove the candidate at row 5, column 5, number 6 from the logic. We then look at the remaining logic to see if any rank 1 or lower regions exist (so as to at least match the rank 1 elimination from the case where the triplet was occupied). The logic now looks like:
The sub-collection consisting of two truth-sets (6R2, 6B5) and three link-sets (6C5, 6C9, 6R6) is a rank 1 sub-region. So, again, the candidate at row 6, column 9, number 6 is eliminated by rank 1 logic. All up, regardless of the state of the triplet, the candidate at row 6, column 9, number 6 is eliminated by rank 1 logic.
The final two cases have multiple triplets. These two cases are currently under discussion in the no-fish thread, as they are template eliminations with no (classical) fish explanation. These eliminations are explainable with Xsudo logic however.
5) No-fish #1
We have 6 truth-sets, 8 link-sets, 1 truth-set triplet (R9C1) and 2 link-set triplets (R8C3, R9C3). As there are 3 triplets, there are 2^3 or 8 possible triplet states. However, with common sense, we won’t have to check all 8 states, as all triplets share a box, so at most one of them is true.
Case I: The truth-set triplet is occupied, and the two link-set triplets are unoccupied.
This case is easy. Since the truth-set triplet is occupied, the truth-set triplet’s link-set is occupied, hence the link-set in row 9 is rank 0.
Case II: The link-set triplet at R8C3 is occupied, and the two others are unoccupied.
This case is also straight-forward. The occupied triplet’s two link-sets are rank 0. There are 5 truth-sets and 6 link-sets remaining. So each of the remaining link-sets is rank 1. Of interest to us is that the link-set in row 9 is rank 1.
Case III: The link-set triplet at R9C3 is occupied, and the two others are unoccupied.
This case is similar to case II. The occupied triplet’s two link-sets are rank 0. There are 5 truth-sets and 6 link-sets remaining. So each of the remaining link-sets is rank 1. Of interest to us is that the link-set in row 9 is rank 0.
Case IV: All triplets are unoccupied.
In this final case, we reduce the logic following the usual procedure. The logic is redrawn with all triplets unoccupied:
We look for a rank 1 or lower sub-region to at least match our worst case so far which is rank 1. Consider the sub-region with one truth-set (9B7) and one link-set (9R9). This is a rank 0 sub-region, which is good enough for us today.
All up, regardless of the triplets’ state, the worst case rank for the link-set in row 9 is rank 1, so the candidate at row 9, column 8, number 9 is eliminated by (reduced) rank 1 logic.
6) No-fish #2
We have 5 truth-sets, 7 link-sets, 1 truth-set triplet (R2C6) and 2 link-set triplets (R2C1, R2C5). As there are 3 triplets, there are 2^3 or 8 possible triplet states. However, with common sense, we won’t have to check all 8 states, as all triplets share a row, so at most one of them is true.
Case I: The link-set triplet at R2C1 is occupied, and the other two triplets are unoccupied.
The occupied triplet’s two link-sets are rank 0. There are 4 truth-sets and 5 link-sets remaining. So each of the remaining link-sets is rank 1. Of interest to us is that the link-set in row 9 is rank 1.
Case II: The link-set triplet at R2C5 is occupied, and the other two triplets are unoccupied.
The occupied triplet’s two link-sets are rank 0. There are 4 truth-sets and 5 link-sets remaining. So each of the remaining link-sets is rank 1. Of interest to us is that the link-set in row 9 is rank 1.
Case III: The truth-set triplet is occupied, and the other two triplets are unoccupied.
We re-draw the logic after setting the triplet’s to the assumed states.
We look for a sub-region that is rank 1 or lower, to at least match the worst case so far which is rank 1. The region with one truth-set (9C1) and 1 link-set (9R9) is rank 0.
Case IV: All triplets are unoccupied:
We re-draw the logic after setting the triplet’s to the assumed states.
We look for a sub-region that is rank 1 or lower, to at least match the worst case so far which is rank 1. The region with one set (9C1) and 1 link-set (9R9) is rank 0.
All up, regardless of the triplets’ state, the worst case rank for the link-set in row 9 is rank 1, so the candidate at row 9, column 5, number 7 is eliminated by (reduced) rank 1 logic.
Virtual Sets:
Hidden Text: Show
Xsudo allows the user to specify one virtual truth-set. This is useful for two solving techniques: Guardians and AURs. We now look at two examples of Virtual X-wings.
Example 1: Guardians
The virtual truth-set (black) consists of two elements; row 2, column 2, number 7 and row 9, column 6, number 7. At least one of these elements must be true to prevent the illegal conjugate loop of length 5 for the digits 7 in cells 4N2,4N4,7N4,8N5 and 8N2. The remaining logic works the same; 2 truth-sets, 2 link-sets, no triplets, so rank 0 logic.
Example 2: AURs (Assumes puzzle has at most one solution)
Use of this technique is an individual’s personal preference; some are comfortable with it, others are not. For those who are happy to use it, here’s how Xsudo does it.
The virtual truth-set (black) consists of two elements; row 4, column 1, number 3 and row 8, column 2, number 3. If the puzzle is assumed to have at most one solution, then at least one of these elements must be true to prevent the deadly pattern in cells 4N1, 4N2, 8N1, 8N2. The remaining logic works the same; 2 truth-sets, 2 link-sets, no triplets, so rank 0 logic.
Example 1: Guardians
The virtual truth-set (black) consists of two elements; row 2, column 2, number 7 and row 9, column 6, number 7. At least one of these elements must be true to prevent the illegal conjugate loop of length 5 for the digits 7 in cells 4N2,4N4,7N4,8N5 and 8N2. The remaining logic works the same; 2 truth-sets, 2 link-sets, no triplets, so rank 0 logic.
Example 2: AURs (Assumes puzzle has at most one solution)
Use of this technique is an individual’s personal preference; some are comfortable with it, others are not. For those who are happy to use it, here’s how Xsudo does it.
The virtual truth-set (black) consists of two elements; row 4, column 1, number 3 and row 8, column 2, number 3. If the puzzle is assumed to have at most one solution, then at least one of these elements must be true to prevent the deadly pattern in cells 4N1, 4N2, 8N1, 8N2. The remaining logic works the same; 2 truth-sets, 2 link-sets, no triplets, so rank 0 logic.
Cannibalism:
Hidden Text: Show
Cannibalism can almost always be explained by using simpler non-cannibalistic logic. It occurs when the assumption of a link-set triplet being occupied results in a negative rank sub-region in the puzzle. The link-set triplet must thus be unoccupied, so a truth-set digit is eliminated (cannibalised).
Example 1: Mutant Swordfish
We have 3 truth-sets, 3 link-sets and 1 link-set triplet at row 1, column 4, number 5. Consider the case where the link-set triplet is occupied. So, we remove the triplet’s truth-set and two link-sets from the logic:
We are left with a rank -1 sub-region; two truth-sets and one link-set. This has no solution, so the assumption that the triplet was occupied is false, and the candidate may be eliminated (cannibalised).
In this example, what really happened was that we missed the (simpler) empty rectangle:
Two truth-sets, three link-sets, no triplets, so nice simple rank 1 logic.
Example 2:
What a way to finish the guide with arguably the most beautiful piece of logic in all of Sudoku found to date: Champagne’s “Double JExocet”.
We can account for all the eliminations in three steps:
Note that I had to use the complementary finned fish interpretation in the first step, but resorted to the standard interpretation in steps 2 and 3.
So that begs the question, can all Cannibal eliminations be reproduced with non-cannibalistic equivalents???
Example 1: Mutant Swordfish
We have 3 truth-sets, 3 link-sets and 1 link-set triplet at row 1, column 4, number 5. Consider the case where the link-set triplet is occupied. So, we remove the triplet’s truth-set and two link-sets from the logic:
We are left with a rank -1 sub-region; two truth-sets and one link-set. This has no solution, so the assumption that the triplet was occupied is false, and the candidate may be eliminated (cannibalised).
In this example, what really happened was that we missed the (simpler) empty rectangle:
Two truth-sets, three link-sets, no triplets, so nice simple rank 1 logic.
Example 2:
What a way to finish the guide with arguably the most beautiful piece of logic in all of Sudoku found to date: Champagne’s “Double JExocet”.
We can account for all the eliminations in three steps:
Note that I had to use the complementary finned fish interpretation in the first step, but resorted to the standard interpretation in steps 2 and 3.
So that begs the question, can all Cannibal eliminations be reproduced with non-cannibalistic equivalents???
Thanks for the detailed explanation...that clears up a lot of issues for me!
)