Solving Tips

[Chain Reasoning③] Applications: Pattern Classification and Advanced Structures

2025-06-06 · 18 min read
Knowledge Base / Techniques Index / Chain Application Patterns

In the previous two articles, we learned about the concepts of strong links and weak links and chain building and propagation rules. This article will systematically introduce various application patterns of chain reasoning and demonstrate how to understand specific techniques using a unified chain framework.

Chain Reasoning Series (3/3)
← ① Basics ← ② Building ③ Applications (Current)
Congratulations! This is the final article in the series
Chain classification diagram
Chain structure classification system: categorized by form, content, and complexity

Classification by Form: Open Chains vs Closed Chains

Based on whether the chain's endpoints connect, chains can be classified as open chains and closed chains (loops).

Open Chain

Characteristics of Open Chains:
  • The chain has a clear start and end point
  • The endpoints do not connect
  • Conclusions are based on the relationship between endpoints

Open chains are the most common chain structure. When the two ends of a chain have a weak link relationship (can see each other), candidate elimination can be performed.

Ex AIC Open Chain:A ═ B - C ═ D - E ═ F
If A and F can see each other (weak link exists), then one of A or F must be true, allowing elimination of other same-digit candidates that can see both A and F.

Closed Chain / Loop

Characteristics of Closed Chains:
  • The chain's endpoint connects back to the starting point, forming a loop
  • Can be used to directly determine the truth value of certain candidates
  • The parity of the loop determines the conclusion type

Closed chains can be divided into continuous loops (Nice Loop) and discontinuous loops based on their structure.

Continuous Continuous Loop:Links in the loop strictly alternate, allowing infinite cyclic tracking
All nodes in the loop can be divided into two color groups: same color means same truth value, different colors mean opposite.
Discontinuous Discontinuous Loop:The loop has consecutive links of the same type at some point, creating a contradiction during tracking
The candidate at the contradiction point can be determined to be true or false.

Classification by Content: Single-digit Chains vs Bi-value Chains

Based on the type of candidates in the chain, chains can be classified as single-digit chains and bi-value chains.

Single-digit Chain

All nodes in the chain are candidates of the same digit. Links come from conjugate pairs (only two positions in a unit have this digit).

Features
  • Tracks only one digit's relationships across different positions
  • Strong links come from conjugate pairs
  • Weak links come from other positions in the same unit
  • Representative techniques: X-Wing, Skyscraper, X-Chain
Single-digit chain diagram
Single-digit chain: tracking conjugate pair relationships of the same digit across positions

Bi-value Chain (XY-Chain)

All nodes in the chain come from bi-value cells (cells with only two candidates). Links transition between different digits.

Features
  • All nodes come from bi-value cells
  • The two candidates within a cell form a strong link
  • Adjacent cells sharing a candidate form a weak link
  • Representative techniques: XY-Wing, XY-Chain, Remote Pairs
The Essence of XY-Chain:
XY-Chain is an alternating chain composed purely of bi-value cells. For example:
R1C1{3,5}(5) - R1C4{5,7}(7) - R3C4{7,9}(9) - R3C8{4,9}(4)
Start is 3, end is 4; candidates 3 and 4 that can see both endpoints can be eliminated.

Mixed Chain (AIC)

The chain contains both single-digit chain nodes and bi-value chain nodes. This is the most versatile chain structure.

Features
  • Flexibly combines various link sources
  • Can freely transition between single-digit and bi-value nodes
  • Most expressive power, can discover more eliminations
  • Representative technique: AIC (Alternating Inference Chain)

Grouped Links

Grouped links treat multiple candidates as a single entity in chain reasoning. This greatly expands the application range of chain techniques.

The Concept of Grouping:
When all candidate positions of a digit within one unit (row/column/box) are concentrated in the intersection area of another unit, these positions can be treated as a "group".

Example: Digit 5 in box 1 appears only in three positions in row 1; these three positions can participate in chains as a group.

Grouped Strong Links

When a group and another candidate/group satisfy the relationship "exactly one is true", a grouped strong link exists.

Ex Digit 5 in box 1 appears only in positions R1C1 and R1C2, forming group A.
In row 1's other positions (box 2 and box 3), digit 5 appears only in position R1C8, as single point B.

A strong link exists between group A and B: row 1 must have one 5, either in group A (box 1) or in B (R1C8).

Grouped Weak Links

When a group and another candidate/group are in the same unit, a grouped weak link exists between them.

Grouped links diagram
Grouped links: multiple candidate positions participating in chain reasoning as a single entity

Discontinuous Loop

A discontinuous loop is a special type of closed chain where a "discontinuity" occurs at some node—the two adjacent links at that node are the same type (both strong links or both weak links).

Types of Discontinuous Loops:
  • Type 1 (Two consecutive strong links):The candidate at the discontinuous point must be false
  • Type 2 (Two consecutive weak links):The candidate at the discontinuous point must be true

Type 1: Two Consecutive Strong Links

Analysis Loop form: A ═ B - C ═ D - ... ═ A (returns to start with strong link)

Assume A is false:
→ through loop propagation → A is true (contradiction!)

Assume A is true:
→ the other end of the last strong link (let's call it X) can be true or false → no contradiction

However, if we track "false" starting from X:
X false → A true (strong link) → ... → X true
This means X cannot be false, so X is true, therefore A is false.

Conclusion: The discontinuous point A must be false.

Type 2: Two Consecutive Weak Links

Analysis Loop form: A - B ═ C - D ═ ... - A (returns to start with weak link)

Assume A is true:
→ through loop propagation → A is false (contradiction!)

Conclusion: The discontinuous point A must be false... wait, that doesn't seem right?

Actually, for Type 2, we need to analyze more carefully. The correct conclusion is:
If tracking "true" from A eventually returns to A requiring A to be false, this creates a contradiction.
Conclusion: The discontinuous point A must be true.

Understanding Common Techniques Through Chain Reasoning

Many seemingly different Sudoku techniques can be understood uniformly through the chain reasoning framework.

Technique Name Chain Description Chain Characteristics
X-Wing 4-node single-digit chain loop Conjugate pairs in 2 rows and 2 columns forming rectangle
Skyscraper 4-node single-digit open chain Two conjugate pairs sharing one end
2-String Kite 4-node single-digit open chain Row and column conjugate pairs connected through box
XY-Wing 3-node bi-value chain Pivot connecting two wings
XY-Chain Multi-node bi-value chain Pure bi-value cell chain
Remote Pairs Even-node bi-value chain Bi-value cell chain with same candidates
W-Wing Mixed chain Bi-value cells connected through conjugate pair
AIC General mixed chain Alternating chain with any combination

Chain Technique Selection Strategies

In actual solving, how do you choose the appropriate chain technique? Here are some suggestions:

1 Simple before Complex:
Start with simple techniques like conjugate pair reasoning and Skyscraper, then try complex AICs.
2 Focus on Bi-value Cells:
Bi-value cells are excellent materials for building chains. When there are many bi-value cells, prioritize XY-Wing and XY-Chain.
3 Look for Conjugate Pairs:
For a digit that's difficult to eliminate, check if it forms conjugate pairs in various units—you might discover single-digit chains.
4 Goal-oriented:
If you want to eliminate a specific candidate, try building a chain where both ends can "see" that candidate.

The Value of Chain Reasoning

The value of learning chain reasoning theory lies not only in being able to use more advanced techniques, but also in:

Advantages of Chain Thinking:
  • Unified Understanding:Understand numerous specific techniques with one framework
  • Flexible Application:Not bound by fixed patterns, flexibly build chains according to the situation
  • Discover New Chains:Don't rely on memorizing specific patterns, but discover them by understanding principles
  • Deep Understanding of Sudoku:Understand the relationships between candidates from a logical essence

Summary

Through these three articles, we have systematically learned the theoretical foundation of chain reasoning:

  • First Article:Definitions, sources, and properties of strong links and weak links
  • Second Article:Chain building rules, propagation logic, and coloring concepts
  • Third Article:Chain classification, application patterns, and unified understanding of common techniques

After mastering these theories, you possess the ability to understand and discover various chain techniques. Through continuous application and consolidation in practice, chain reasoning will become a powerful tool for solving complex Sudoku puzzles.

Practice Now:
Start a Sudoku game, and try analyzing candidate relationships with chain thinking! When you encounter difficulties, think:
  • Where are the bi-value cells? Can they form chains?
  • In which units does a certain digit form conjugate pairs?
  • Can I find a chain where both ends see the candidate I want to eliminate?