Sudoku Chains

A Sudoku Chain is a sequence of candidate inferences in which the truth or falsity of one state forces the next state.

The core language is built from two relationships:

• Strong Link: false → true;
• Weak Link: true → false.

When those inference roles alternate, a logical state can travel across the grid:

false → true → false → true → ...

That propagation can prove:

• a candidate must be true;
• a candidate must be false;
• at least one of two endpoints must be true;
• an assumption leads to a contradiction;
• or multiple possible starts lead to the same conclusion.

Chains are not about drawing long lines across a Sudoku for their own sake.

The purpose is to find the shortest verifiable sequence that proves a useful deduction.

Before Chains: understand candidate states

A chain does not connect whole cells in a vague geometric sense.

It connects candidate states.

For example:

r4c2=7

is one logical state: “candidate 7 in r4c2 is true.”

A link explains what would happen to another candidate state if that first state were true or false.

This matters because the same cell can participate in several different links through different candidate values.

When chain solving feels confusing, return to the individual candidate states and state every inference aloud.

The two inference directions

Strong Link

A false → B true

A and B cannot both be false.

Weak Link

A true → B false

A and B cannot both be true.

The alternation works because each link produces the state the next link needs.

A Strong Link takes false and produces true.

A Weak Link takes true and produces false.

Then another Strong Link can continue the process.

A four-node chain

Suppose:

A = B - C = D

where:

• = represents a Strong inference role;
• - represents a Weak inference role.

Start with A false.

A false

The Strong Link forces:

B true.

B true

The Weak Link forces:

C false.

C false

The next Strong Link forces:

D true.

So:

A false → D true.

If the same alternating structure can also be read from the other end:

D false → A true.

then A and D cannot both be false.

Therefore:

At least one endpoint is true.

Any candidate state that conflicts with both possible true endpoints can be eliminated.

Why a chain can often be read in both directions

Ordinary conjugate pairs and bivalue-cell pairs are both Strong and Weak relationships.

That allows a valid alternating sequence to be followed from either endpoint when the link roles are assigned correctly.

For an endpoint-elimination chain, the useful proof is usually:

Endpoint A false → Endpoint B true
Endpoint B false → Endpoint A true

Therefore both endpoints cannot be false.

This is the logical basis of the standard common-peer eliminations taught in X-Chain and XY-Chain.

Do not assume every arbitrary sequence of candidate connections has this reversible endpoint property. The link type and direction must be valid at every step.

Chains are not arbitrary guessing

Chain proofs often begin with a temporary statement such as:

Suppose candidate A is false.

That can sound like guessing, but the role of the assumption is different.

You do not accept the assumption as a solution.

Instead:

1. make one temporary state;
2. follow only forced implications;
3. reach a consequence;
4. repeat from the opposite endpoint or alternative state when required;
5. keep only the conclusion that is logically guaranteed.

The proof is in the forced links, not in the initial temporary premise.

This is closer to a mathematical proof by cases or contradiction than to unsupported trial and error.

Named Sudoku techniques are compressed chains

Many patterns that are easier to recognize geometrically can also be described as short chains.

Skyscraper

A short one-digit alternating path between two conjugate-pair structures.

Two-String Kite

A short one-digit chain whose Weak connection is created through a box.

W-Wing

Two matching bivalue cells interact through an external Strong Link.

XY-Wing

A three-cell bivalue propagation pattern that can be represented as a short XY-Chain.

The named pattern is usually the better human description when the geometry is obvious.

Chain language becomes valuable when:

• the pattern stretches farther;
• the geometry no longer matches one familiar shape;
• or you want one logical framework that explains several techniques.

Links inside and between cells

Chains can move through different kinds of candidate relationships.

Inside a bivalue cell

Cell:

{3,8}

The candidates are strongly linked:

not 3 → 8

and:

not 8 → 3.

Because a cell holds only one digit, they are weakly linked as well.

Between peer cells

Two candidate 8s that see each other are weakly linked:

first 8 true → second 8 false.

Within a conjugate pair

Two same-digit candidates that are the only positions in one house are both Strong and Weak.

These three ideas are enough to understand the current X-Chain and XY-Chain Guides.

One-digit Chains

An X-Chain follows one candidate digit only.

For example, every node may represent candidate 7 in a different cell.

Strong Links normally come from:

• row conjugate pairs;
• column conjugate pairs;
• box conjugate pairs.

Weak Links come from candidate 7 positions that see each other.

A standard eliminative X-Chain starts and ends with a Strong inference role. That proves at least one endpoint contains the chain digit, so the digit can be eliminated from common peers of the endpoints.

Short X-Chains overlap with named single-digit patterns such as Skyscraper, Two-String Kite, Turbot Fish, and Empty Rectangle.

Bivalue-cell Chains

An XY-Chain uses only bivalue cells in its standard form.

Inside each chain cell:

• the two candidates provide the Strong step.

Between consecutive cells:

• the shared candidate provides a Weak step because the cells see each other.

The candidate value changes as the chain travels.

The first and final cells contain the same endpoint candidate Z. This proves that at least one endpoint is Z, allowing Z to be eliminated from cells that see both endpoints.

XY-Wing is the short introductory example of this style of propagation.

Coloring is another way to manage chain states

Simple Coloring follows one digit through a connected network of conjugate pairs.

Instead of tracing one chosen path, it assigns two alternating states — often shown as Color A and Color B — to the whole connected Strong-Link component.

Then it searches for contradictions or outside candidates that interact with both polarities.

So Coloring and Chains are not rival theories.

They are different human interfaces for closely related binary inference logic.

What makes a chain useful?

A chain may be valid but produce no solving progress.

That is important.

The grid does not reward chain length; it rewards conclusions.

Common useful outcomes include:

Endpoint elimination

At least one of two endpoints is true.

A candidate seeing both endpoints cannot be true.

Contradiction

A premise eventually forces an impossible state, proving the premise false.

Forced placement

One branch or endpoint assumption forces the same candidate true under every possibility.

Convergence

Different possible starting states lead to the same result.

If your sequence reaches no guaranteed consequence, it may be a valid implication path but it is not yet a productive Sudoku move.

Open chains, loops, and the road to AIC

The Guides currently focus on open chains with useful endpoint conclusions.

More general chain theory includes:

• Alternating Inference Chains (AICs);
• discontinuous loops;
• continuous loops;
• grouped nodes;
• ALS nodes.

Those extensions allow more varied endpoint conditions and eliminations.

They belong to the Expansion layer because the current corpus first needs a stable foundation in candidate-level links, X-Chains, XY-Chains, and Coloring.

The key idea will not change:

alternate valid inference roles and preserve the forced logic at every step.

Chain vs Net

A chain progresses linearly: each new inference follows from the state immediately before it.

A Net branches or requires the combined result of more than one earlier branch.

That distinction matters because a net is harder to verify manually and belongs to more advanced forcing methods.

For the current learning path:

Strong Links
→ Weak Links
→ short Chains
→ Coloring / X-Chain / XY-Chain
→ AIC
→ grouped and forcing structures later

Chain notation

Notation is useful only after the logic is understood.

A simple conceptual notation is:

A = B - C = D

where:

• = means a Strong inference role;
• - means a Weak inference role.

Sudoku communities also use Nice Loop notation, AIC/Eureka notation, candidate coordinates, digit labels, and grouped-node syntax.

Those systems compress a chain so it can be communicated precisely.

They do not change the proof.

When learning, always be able to expand notation back into plain language:

If this is false, that is true; therefore this is false; therefore that is true...

How to verify a chain

Before making an elimination, check every step.

1. Identify the exact candidate node.
2. Name the link source. Row, column, box, bivalue cell, or another valid structure.
3. State the inference direction. False → true or true → false.
4. Continue only if the next state is forced.
5. Verify the endpoint conclusion in both required cases.
6. Check the target actually conflicts with every endpoint possibility used by the proof.

If one step relies on “probably” or “looks connected,” the chain is not proven.

How to search for Chains without getting lost

Do not start by trying to enumerate the entire candidate graph.

A practical progression is:

1. find a promising Strong Link;
2. inspect short Weak connections from its endpoints;
3. look for another Strong Link;
4. stop as soon as a useful endpoint relationship appears;
5. prefer a named short pattern when one describes the same proof clearly;
6. after any elimination, return to Singles and simpler techniques.

This keeps Chains as a solving tool rather than a graph-drawing exercise.

Common mistakes

Alternating line styles without checking the inference

The link must be logically valid. A solid or dashed line does not make it so.

Reading a Weak Link backward

From A false, a weak-only relationship gives no forced conclusion about B.

Forgetting that candidate states, not just cells, are the nodes

The same cell may participate through different candidates.

Building a chain with no conclusion

A long valid path can still be useless.

Eliminating a target that conflicts with only one endpoint

For the standard endpoint proof, the target must be incompatible with every possible true endpoint.

Jumping straight into formal notation

Notation should compress reasoning you already understand.

Using an unnecessarily long chain

If a Skyscraper, W-Wing, X-Wing, or other simpler technique proves the same elimination, prefer the shorter explanation.

How to learn Chains without getting overwhelmed

A good progression is:

1. master Strong Links;
2. master Weak Links;
3. trace abstract four-node chains;
4. revisit Skyscraper, Two-String Kite, and W-Wing as chain examples;
5. learn Simple Coloring as a visual binary network;
6. learn short X-Chains;
7. learn short XY-Chains;
8. only then generalize into AICs.

FAQ

What is a Sudoku Chain?

A Sudoku Chain is a sequence of forced candidate implications. Each state forces the next through Strong and Weak relationships until the sequence proves a useful conclusion.

Why do Chains alternate Strong and Weak links?

Strong Links turn false into true, while Weak Links turn true into false. Alternating those inference roles lets a forced state propagate.

Are Chains the same as guessing?

No. A temporary assumption may start a proof, but every following state is forced and only a guaranteed conclusion is kept.

Are all Chains expert techniques?

No. Several familiar Advanced patterns are short chains in disguise. The difficulty comes from recognizing and verifying longer or more general structures.

Can XY-Wing be represented as a chain?

Yes. It can be represented as a short XY-Chain through three bivalue cells.

What is an AIC?

An Alternating Inference Chain is the general framework that alternates Strong and Weak inference roles. A dedicated AIC Guide is planned for the Expansion layer.

What is the difference between a Chain and a Net?

A chain is linear: each step depends directly on the previous one. A net branches or combines implications from multiple earlier branches.

What to learn next

Study Simple Coloring, X-Chain, and XY-Chain. Together they turn the abstract Strong/Weak language into three practical solving methods before the later AIC expansion.