Module 4 · Agniveer Reasoning

Puzzles

Updated 18 Nov 2026 · 8,558 words · 4 sections · illustrated
Updated 21 Nov 2026
Seating, comparison puzzles.
Seating Arrangement · Comparison
🎯 Take Mock Test
Linear SeatingArrangementCircular SeatingArrangementDouble Row &ComplexArrangementsSequencing PuzzlesRanking PuzzlesSeating ArrangementsComparison PuzzlesExam Hook: Seating puzzles test directional logic and neighbour clues.Exam Hook: Comparison puzzles test relative ordering and attribute mapping.vlymbooq
Puzzle Types Comparison for GS1 Geography — Compares seating and comparison puzzle types to highlight key solving strategies for exam. · Open the full diagram →

Learning Objectives

  • Understand the fundamental concepts of Puzzles
  • Apply key formulas and techniques to solve problems
  • Practice with exam-level questions to build speed and accuracy

1. Linear Seating Arrangement

1.1 Basic Principles of Linear Arrangement

Linear seating arrangement problems present a row of positions that are ordered from left to right, and the solver must place each person or object into a unique slot using the given clues. The fundamental principle is that every clue either fixes a position absolutely, relates two positions by a known offset, or excludes certain slots for a particular entity. Because the row has two distinct ends, the concepts of leftmost and rightmost are absolute references that never change regardless of the perspective of the individuals seated. This absolute frame of reference makes linear puzzles the most straightforward type of arrangement, yet the difficulty rises quickly when clues are indirect, such as “A sits third to the left of B” combined with “C is not at either end”. The solver must translate each verbal statement into a precise positional constraint, often representing the row as a numbered line of slots from 1 to N. The process of solving begins with identifying the most restrictive clues — those that mention ends, exact distances, or immediate neighbours — because they reduce the number of possible configurations dramatically. Once a few definite positions are established, the remaining relative clues can be applied iteratively, and any contradiction forces the solver to backtrack and try an alternative placement. Mastery of linear arrangement therefore rests on systematic notation, careful counting of offsets, and the discipline to use only the information explicitly given.

1.2 Definite and Relative Position Clues

Clues in linear puzzles fall into three broad categories: definite, relative, and negative. A definite clue pins an entity to a specific slot, for example “D sits at the left end” or “E occupies the fourth position from the left”. Such clues are the anchors of the arrangement because they require no further inference. Relative clues describe the positional relationship between two entities without fixing either absolutely; typical forms include “A is second to the left of B”, “C sits immediately to the right of D”, or “there are exactly two persons between F and G”. These clues create a set of possible pairs of slots that satisfy the stated offset, and the solver must keep all viable pairs until other clues eliminate the impossible ones. Negative clues state where an entity cannot be, such as “H is not at either end” or “I does not sit next to J”. Although they do not place anyone directly, negative clues are powerful because they prune the search space and often resolve ambiguities left by relative clues. The interplay of these three clue types is the engine of the puzzle: a definite clue may satisfy a relative clue automatically, a relative clue may narrow the options for a negative clue, and a negative clue may force a unique placement for a relative pair. Recognising which clue type each statement belongs to, and processing them in order of restrictiveness, is the hallmark of an efficient solving strategy.

1.3 Negative Conditions and Elimination

Negative conditions are frequently the key to breaking deadlocks in linear arrangements. When a puzzle provides several relative clues that each allow multiple configurations, the solver can draw a table of possible slots for each person and then systematically cross out slots prohibited by negative statements. For instance, if “K is not at an end” and “K is not immediately left of L”, the slots 1 and N are removed from K’s list, and any slot that would place K immediately left of L is also removed. This elimination process is greatly aided by a visual grid where rows represent persons and columns represent positions; each clue adds marks or crosses, and the goal is to reach a state where every row has exactly one possible column. The method mirrors the logic of a Latin square and ensures that no hidden assumption creeps in. It is important to remember that a negative clue about a pair, such as “M and N are not adjacent”, eliminates two specific ordered pairs (M left of N and N left of M) rather than removing a single slot. As the grid fills, the remaining empty cells often reveal forced placements: if only one person can occupy slot 3, that person must go there, and the corresponding row and column are then cleared for others. This cascade effect is the primary mechanism by which a seemingly under‑determined puzzle becomes fully determined.

1.4 Multiple Row Extensions

Some linear puzzles extend the single row into two or more parallel rows, each with its own left‑to‑right orientation, and introduce clues that link positions across rows. The most common variant is the double‑row arrangement where one row faces the other, so that the person in the first slot of the upper row sits directly opposite the person in the first slot of the lower row. In such puzzles, the solver must maintain two independent linear sequences while also tracking the cross‑row relationships such as “P sits opposite Q” or “R is diagonally opposite S”. The same three clue categories apply within each row, but additional cross‑row clues create a second dimension of constraints. A useful technique is to draw the two rows one above the other, aligning the columns so that opposite seats share the same column index. Then each clue can be translated into a mark on this two‑dimensional grid. The solving order remains: place definite clues first, then apply relative clues within each row, then enforce cross‑row conditions, and finally use negative clues to eliminate remaining ambiguities. Because the number of variables doubles, the grid becomes larger, but the logical structure is identical to the single‑row case; the only extra demand is careful bookkeeping to avoid mixing up the two rows.

Comparison of Linear, Circular, and Double Row Arrangement Features
FeatureLinear ArrangementCircular ArrangementDouble Row Arrangement
Reference FrameFixed left and right endsNo absolute ends; positions defined by clockwise/anticlockwise orderTwo parallel linear frames with fixed ends
Direction of Left/RightAbsolute for all personsDepends on facing direction (centre or outward)Absolute within each row; opposite seats share column
Neighbour DefinitionImmediate left/right in the rowImmediate clockwise/anticlockwiseImmediate left/right within row; opposite across rows
Typical Clue TypesEnd positions, exact offsets, non‑adjacencyOpposite, immediate left/right, between, facing directionOpposite, diagonal, same column, row‑specific offsets
Complexity DriversNumber of persons, indirect relative cluesFacing direction reversals, multiple circlesCross‑row constraints, unequal row lengths

2. Circular Seating Arrangement

2.1 Orientation and Direction Conventions

Circular seating arrangement problems place persons around a round table where the only structural references are the order of seats around the circumference. Unlike a line, a circle has no ends, so the concepts of left and right are not absolute but depend entirely on the direction each person is facing. The standard convention in reasoning tests is that “left” means the immediate neighbour in the anticlockwise direction when the person faces the centre, and “right” means the immediate neighbour in the clockwise direction. If a person faces outward, the meanings of left and right are reversed: left becomes clockwise and right becomes anticlockwise. This reversal is a frequent source of error, and the solver must note the facing direction of every individual before interpreting any left/right clue. The circle is usually drawn with a fixed number of equally spaced positions, often numbered 1 to N in clockwise order for the solver’s convenience, but the numbering is only a mental aid; the actual puzzle never refers to absolute numbers. The first step in solving is to decide on a reference point — often the position of a person mentioned in a definite clue such as “P sits at the head” (which in a circle simply means we can place P arbitrarily and treat that seat as position 1). All other placements are then made relative to this anchor.

2.2 Fixed and Relative Positions on a Circle

Clues in circular puzzles mirror those in linear puzzles but are expressed in circular terminology. A definite clue might state “Q sits exactly opposite P”, which in an even‑sized circle fixes Q to the seat N/2 steps clockwise from P. A relative clue such as “R is second to the left of S” translates to a fixed offset of two steps anticlockwise from S if S faces the centre, or two steps clockwise if S faces outward. The phrase “between” also appears: “T sits between U and V” means that moving clockwise from U, the next seat is T, and the next after that is V (or the reverse order). Because the circle is closed, “between” always implies two possible orders unless the direction is specified. The solver must therefore keep both orders as separate cases until other clues eliminate one. The most powerful clues are those that combine a fixed reference with a relative offset, for example “W is third to the right of the person who sits opposite X”. Such a clue chains two relationships and often collapses many possibilities at once. As with linear puzzles, the solver should first place all definite clues, then apply relative clues to generate possible positions, and finally use negative clues to discard impossible cases.

2.3 Facing Inside Versus Facing Outside

The facing direction of each person is a critical attribute that changes the interpretation of every left/right clue involving that person. In many puzzles, all persons face the centre, which simplifies the convention: left is always anticlockwise, right is always clockwise. However, puzzles often mix facing directions — some face the centre, some face outward — and may even include a clue such as “A faces the centre while B faces outward”. When directions differ, the solver must maintain a separate facing flag for each person. A useful technique is to draw the circle with small arrows indicating the facing direction of each seated person. Then, when a clue says “C is immediate left of D”, the solver looks at D’s arrow: if D faces centre, immediate left is the anticlockwise neighbour; if D faces outward, immediate left is the clockwise neighbour. This visual aid prevents the common mistake of applying a uniform left/right rule to all persons. Moreover, facing direction can itself be the subject of a clue, for example “E faces the same direction as F” or “G faces opposite to H”. These clues add another layer of logical constraints that must be satisfied simultaneously with the positional clues. The interplay between position and orientation is what makes circular puzzles richer than linear ones, and a systematic approach that tracks both attributes in parallel is essential for accurate solving.

2.4 Complex Circular Puzzles with Multiple Attributes

Advanced circular puzzles attach additional attributes to each person — such as profession, colour preference, or city of residence — and provide clues that link these attributes to positions or to each other. For instance, “The doctor sits third to the left of the person who likes red” combines a positional offset with an attribute association. The standard method for such puzzles is to extend the circle diagram into a table where each seat has columns for the person’s name and each attribute. Definite clues fill cells directly; relative clues create links between seats; attribute clues create links between attribute columns. The solver then uses a process of constraint propagation: when a seat’s person is known, all attribute clues involving that person become clues about the attribute columns of other seats. This often triggers a cascade that resolves many attributes at once. Because the number of variables multiplies, it is crucial to keep the notation clean and to revisit all clues after each new deduction. The principle remains the same: start with the most restrictive clues, maintain multiple cases when ambiguity persists, and eliminate cases that lead to contradictions. Mastery of multi‑attribute circular puzzles is a strong indicator of overall reasoning proficiency, as it demands simultaneous tracking of spatial, directional, and categorical information.

Worked Example: Circular Arrangement with Mixed Facing Directions
Eight persons A, B, C, D, E, F, G, H sit around a circular table. Four face the centre and four face outward. A faces the centre and sits third to the left of B. C faces outward and sits immediate right of D. E faces the centre and sits opposite F. G sits second to the right of H. H faces outward. B and D face opposite directions. Determine the seating order and facing directions.
Place A at an arbitrary seat (seat 1) facing centre. Since A is third to the left of B and A faces centre, left is anticlockwise; thus B is three seats clockwise from A, so B occupies seat 4. B and D face opposite directions; we do not yet know B’s direction. C faces outward and sits immediate right of D. Immediate right of D depends on D’s facing: if D faces centre, right is clockwise; if D faces outward, right is anticlockwise. E faces centre and sits opposite F; in an 8‑seat circle opposite means four seats apart. G sits second to the right of H; H faces outward, so for H right is anticlockwise, thus G is two seats anticlockwise from H. We now test the two possibilities for D’s facing. Case 1: D faces centre. Then C is immediate right (clockwise) of D. Case 2: D faces outward. Then C is immediate right (anticlockwise) of D. Using the fixed positions of A (1) and B (4) and the opposite pair E/F, we systematically fill the remaining seats. Only Case 2 yields a consistent arrangement without contradiction. The final clockwise order from seat 1 is: A (centre), H (outward), G (centre), B (outward), D (outward), C (outward), E (centre), F (outward). All clues are satisfied.
Exam‑Day Insight
When a circular puzzle mixes facing directions, draw a small arrow on each seat as soon as the direction is known; this turns every left/right clue into a simple clockwise or anticlockwise move without mental reversal, saving precious seconds and preventing the most common error.

3. Double Row and Complex Arrangements

3.1 Two Row Facing Each Other

Double row puzzles present two parallel lines of seats, usually of equal length, with the occupants of one row facing the occupants of the other row. The most natural mental model is to draw the upper row (Row 1) and the lower row (Row 2) such that the first seat of Row 1 is directly opposite the first seat of Row 2, the second opposite the second, and so on. This alignment makes the concept of “opposite” unambiguous: two persons are opposite if they share the same column index. Clues in these puzzles fall into three groups: intra‑row clues that behave exactly like linear arrangement clues (e.g., “P sits second to the left of Q in the upper row”), inter‑row clues that relate positions across rows (e.g., “R sits opposite S”, “T sits diagonally opposite U”), and global clues that involve both rows simultaneously (e.g., “The person who likes cricket sits in the lower row exactly behind the person who likes football”). The solving strategy is to treat each row as an independent linear arrangement first, using only intra‑row clues, and then to overlay the inter‑row constraints. Because the two rows share the same column indices, a placement in one row immediately restricts the opposite seat in the other row. A powerful technique is to maintain a single combined grid with two rows and N columns, marking definite placements, relative offsets within each row, and cross‑row links. Negative clues such as “V does not sit in the upper row” or “W is not opposite X” are applied to this grid exactly as in the single‑row case. The double row format often appears with additional attributes (profession, colour, etc.), turning the grid into a multi‑column table where each cell holds a bundle of attributes. The logical process remains constraint propagation: each new deduction reduces the domain of possible values for other cells until a unique solution emerges.

3.2 Three or More Parallel Rows

Some puzzles extend the double row concept to three or more parallel rows, often representing floors of a building, lanes of a race track, or tiers of a stadium. The fundamental logic does not change: each row is a linear sequence with its own left‑to‑right orientation, and clues may relate positions within a row, between adjacent rows, or even between non‑adjacent rows. The solver must decide on a consistent coordinate system — typically row number (1 for topmost, 2 for next, etc.) and column number (1 for leftmost). A clue like “The person on the third floor sits two seats to the right of the person on the first floor who faces north” translates into a constraint linking row 3, column c+2 to row 1, column c, provided the facing direction on floor 1 makes “right” correspond to increasing column index. When the number of rows grows, the grid becomes larger but the solving algorithm is identical: populate definite clues, propagate relative offsets, enforce cross‑row conditions, and prune with negatives. The main challenge is bookkeeping; a well‑organised spreadsheet‑style diagram with rows as horizontal bands and columns as vertical lines keeps the information manageable. It is also common for such puzzles to include a “view” clue, for example “From the left end of the second row, the third seat of the fourth row is visible”, which imposes a geometric constraint that can be modelled by checking column alignment. Mastery of multi‑row puzzles demonstrates the ability to scale the basic linear logic to higher dimensions without losing rigor.

3.3 Integrating Linear and Circular Logic

Occasionally a puzzle combines a linear arrangement with a circular one, for instance by seating some persons in a row and others around a table, with clues linking the two groups. The solver must maintain two separate spatial models — a line and a circle — and a set of bridge clues that connect them. A typical bridge clue might be “The person who sits at the left end of the row faces the person who sits opposite the centre of the circle”. To handle this, the solver first solves each sub‑puzzle as far as possible independently, using only the clues that are internal to that sub‑puzzle. The bridge clues are then treated as additional constraints that relate variables from the two models. Because the two models have different coordinate systems, the solver must translate each bridge clue into a statement about the variables of both models. For example, “left end of the row” becomes a specific person variable in the linear model; “opposite the centre of the circle” becomes a specific seat variable in the circular model. The bridge clue then asserts an equality or a relation between those two variables. The combined system is solved by iterating deductions in one model, feeding the results into the other model via the bridge clues, and repeating until both models are fully determined. This integration tests the solver’s ability to switch between spatial reasoning frameworks while keeping the logical dependencies clear.

Comparison of Clue Types Across Arrangement Families
Clue CategoryLinearCircularDouble RowMulti‑Row
Definite PositionEnd seat, exact indexOpposite a known person, fixed seat if numberedSpecific row and column, opposite a known personSpecific floor and column
Relative OffsetLeft/right by N seatsClockwise/anticlockwise by N seats (depends on facing)Left/right within row; opposite/diagonal across rowsLeft/right within row; vertical offsets between rows
NegativeNot at end, not adjacentNot opposite, not immediate neighbour, not facing a directionNot in a given row, not opposite, not diagonalNot on a given floor, not in a given column
Attribute LinkPerson‑attribute at a seatPerson‑attribute at a seat (with facing)Person‑attribute in a cell (row, column)Person‑attribute in a cell (floor, column)
Bridge ClueRareRareLinks two rows directlyLinks non‑adjacent rows or row to external model

4. Comparison Puzzles: Sequencing and Ranking

4.1 Temporal Sequencing Using Before and After Clues

Sequencing puzzles ask the solver to order a set of events, tasks, or persons in time or priority based on comparative statements such as “Event A occurs before Event B”, “Event C occurs immediately after Event D”, or “Event E is the third event”. The underlying structure is a total order on a finite set, and each clue provides a binary relation (precedes, follows, immediate precedence, or exact rank). The most direct way to represent the problem is a directed graph where vertices are events and an edge from X to Y means X precedes Y. A clue like “A before B” adds an edge A→B; “C immediately after D” adds edges D→C and also forbids any other event between D and C. “E is third” fixes E’s position in the linear order. The solver’s task is to find a topological ordering of the graph that satisfies all constraints, including the immediate‑precedence constraints which are stricter than simple precedence. When the graph contains a cycle, the clues are contradictory and the puzzle is flawed; in well‑formed puzzles the graph is a DAG (directed acyclic graph) and at least one topological order exists. If multiple orders satisfy all clues, the puzzle usually asks for a specific piece of information that is invariant across all valid orders, such as “Which event must be first?”

Conditional Clues and Logical Dependencies

Many advanced puzzles embed clues that become active only when a specific condition is satisfied. A typical formulation is “If P sits at an end, then Q sits immediately to the right of R”. The solver must treat the antecedent as a hypothesis, explore both the case where it holds and the case where it does not, and propagate consequences in each branch. This creates a binary decision tree that can be pruned using contradictions such as duplicate assignments or violation of a negative clue. The key mechanism is to maintain a set of active constraints that are currently enforced and a set of pending constraints awaiting their trigger. When a definite placement satisfies a trigger, the pending constraint moves to the active set and its consequent is applied immediately. Conversely, if a placement makes the antecedent impossible (for example, P is placed in a middle seat), the conditional clue is permanently deactivated and can be discarded. Efficient solvers record the trigger‑consequent pairs in a separate table so that each new definite assignment can be checked against all pending triggers in O(1) average time. The logical dependency graph formed by these conditionals often reveals clusters of clues that must be resolved together; solving one cluster may unlock several others, dramatically reducing the search space.

Examiners frequently combine conditional clues with mutual exclusivity statements such as “Either A sits at the left end or B sits at the right end, but not both”. This is logically equivalent to an exclusive‑or (XOR) and can be modelled by introducing a Boolean variable for each disjunct and adding the constraint (A_end ⊕ B_end). When the puzzle also contains a conditional that references one of the disjuncts, the solver must propagate the XOR through the conditional branches. A practical technique is to create two parallel scenarios — one assuming A_end true, the other assuming B_end true — and solve each independently. If both scenarios lead to a valid complete arrangement, the puzzle typically asks for a property that is invariant across both, such as “Who sits in the middle seat?”. If only one scenario survives, the other is eliminated and the surviving assignments become definite. This scenario‑splitting method is the backbone of solving puzzles with layered conditionality and is essential for high‑difficulty sets.

Multi‑Attribute Matrix Method

When a puzzle assigns several independent attributes — such as colour, profession, and favourite sport — to each person or seat, a single linear diagram becomes insufficient. The multi‑attribute matrix organises information in a two‑dimensional grid where rows represent entities (persons, seats, floors) and columns represent attributes. Each cell holds a value, a set of possible values, or a marker for “unknown”. The solving algorithm proceeds by iteratively applying three operations: direct assignment from a definite clue, elimination from a negative clue, and cross‑attribute propagation from relational clues that link two attributes (e.g., “The engineer likes cricket”). Cross‑attribute propagation is implemented by intersecting the candidate sets of the two linked attributes for the same entity; if the intersection becomes a singleton, that value is fixed and the elimination cascades to other entities sharing the same attribute value. This process is identical to constraint propagation in a finite‑domain constraint satisfaction problem (CSP) and guarantees convergence to a unique solution when the puzzle is well‑posed.

A practical implementation uses a spreadsheet‑style layout with colour‑coding: green for confirmed, yellow for candidates, red for eliminated. The solver first populates all definite clues, then scans each relational clue to shrink candidate sets. Because each attribute domain is typically small (≤ 8 values), the matrix remains manageable even for 12‑entity puzzles. An advanced twist introduced in recent examinations is the attribute‑dependency clue: “The person who likes hockey is not the doctor”. This creates a binary constraint between two attribute columns that does not involve a specific entity. The matrix handles it by marking the (hockey, doctor) pair as forbidden for every row; during propagation, any row that has hockey as its only remaining sport forces doctor to be removed from that row’s profession candidates, and vice‑versa. Mastery of the matrix method allows the solver to visualise the entire constraint network at a glance and to spot hidden symmetries — such as two rows having identical candidate patterns — that can be merged or used to deduce further eliminations.

Matrix Elimination Technique
Five friends — Arun, Bina, Chetan, Divya, and Esha — each have a distinct profession (Doctor, Engineer, Teacher, Lawyer, Artist), a favourite colour (Red, Blue, Green, Yellow, White), and a preferred sport (Cricket, Football, Hockey, Tennis, Badminton). Clues: 1) The Engineer likes Blue. 2) The Doctor does not like Red. 3) The person who likes Cricket is not the Teacher. 4) Arun is the Lawyer and likes Yellow. 5) Bina likes Green and plays Football. 6) Chetan is not the Artist. 7) The Teacher likes White. 8) The person who plays Hockey is the Engineer. 9) Divya does not play Tennis. 10) Esha is not the Doctor. Determine the profession, colour, and sport for each friend.
Create a 5 × 3 matrix (rows = friends, columns = profession, colour, sport). Initialise each cell with the full domain. Apply direct clues: Row Arun → profession Lawyer, colour Yellow. Row Bina → colour Green, sport Football. Clue 1 links Engineer ↔ Blue; mark Engineer‑Blue pair as required. Clue 8 links Engineer ↔ Hockey; thus Engineer must have both Blue and Hockey. Since each profession appears once, the Engineer row gets profession Engineer, colour Blue, sport Hockey. Clue 7 forces Teacher ↔ White. Clue 2 removes Red from Doctor’s colour candidates. Clue 3 removes Teacher from Cricket’s profession candidates. Clue 6 removes Artist from Chetan. Clue 9 removes Tennis from Divya. Clue 10 removes Doctor from Esha. Propagate: Engineer row is now fully fixed (Engineer, Blue, Hockey). Because Blue and Hockey are used, eliminate them from all other rows. Teacher must have White; the only row without a colour fixed and not using White is Chetan or Divya or Esha. Continue elimination until each row has a singleton in every column. The unique solution: Arun — Lawyer, Yellow, Badminton; Bina — Doctor, Green, Football; Chetan — Teacher, White, Cricket; Divya — Artist, Red, Tennis; Esha — Engineer, Blue, Hockey.
Exam‑Day Strategy
When a multi‑attribute puzzle appears, draw the matrix on the rough sheet before reading all clues. Fill in direct assignments immediately; they anchor the grid and reduce the candidate space for every subsequent clue. Keep a separate list of relational clues (links between two attributes) and process them after the first pass, because they often resolve the remaining ambiguities in a single sweep.

Hybrid Puzzles with Quantitative Constraints

Hybrid puzzles combine spatial arrangement with numerical conditions such as “The sum of the floor numbers of the Doctor and the Engineer is 7” or “The person sitting third from the left earns twice the salary of the person at the right end”. These clues introduce arithmetic relations that must be satisfied simultaneously with the positional logic. The solver treats each quantitative clue as an equation or inequality over the domain of the relevant attribute (floor number, salary, age, etc.). Because the domains are discrete and small, the equations can be solved by enumerating all feasible pairs or triples that satisfy the arithmetic condition and then intersecting those feasible sets with the positional candidates. For example, if floors are numbered 1‑5 and the sum of two specific persons’ floors is 7, the only possible unordered pairs are (2,5) and (3,4). This reduces the candidate floors for those two persons to {2,3,4,5} with the additional restriction that they cannot both take 2 or both take 5. The solver records these arithmetic constraints in a separate numeric constraint table and runs a propagation loop: whenever a floor becomes definite for a person, the corresponding partner’s candidate set is filtered to values that keep the sum valid. If the partner’s set becomes a singleton, that value is fixed and the loop continues. This technique is essentially unit propagation on a bipartite graph of arithmetic constraints and works efficiently because each numeric clue involves at most three variables.

Examiners often embed a hidden ordering within the numbers, such as “Salaries are distinct integers from 10 k to 50 k in steps of 10 k”. The distinctness condition turns the numeric domain into a permutation problem. The solver can combine the permutation logic with the arithmetic constraints by maintaining a global “used values” set. When a salary is assigned, it is removed from the pool for all other persons. This interaction between permutation and arithmetic is the hallmark of the most difficult hybrid puzzles. A practical shortcut is to first solve the pure spatial sub‑puzzle ignoring numbers, obtaining a partial order of persons. Then map the numeric domain onto that order using the arithmetic clues as anchors. For instance, if the spatial order is A < B < C < D < E (left to right) and a clue says “The salary of the person at the left end is 10 k”, the entire salary sequence may be forced if another clue gives a fixed difference. Recognising when the spatial order fully determines the numeric assignment saves enormous time.

Optimization and Minimal Information Puzzles

A growing category of puzzles asks not for a complete arrangement but for the minimum number of clues required to uniquely determine a specific answer, or the maximum number of persons that can satisfy a given set of conditions. These are optimisation problems disguised as logical puzzles. The solver must reason about the information content of each clue: a definite position clue (e.g., “A sits at seat 3”) contributes log₂ N bits of information, while a relative clue (“A sits two seats left of B”) contributes less because it leaves a symmetry. By constructing the constraint hypergraph where vertices are variables and hyperedges are clues, one can compute the rank of the constraint matrix over GF(2) for binary clues or over integers for arithmetic clues. In practice, the aspirant does not perform linear algebra but uses a heuristic: identify the critical clues whose removal creates at least two distinct valid arrangements. This is done by tentatively deleting each clue, checking for uniqueness (often by a quick mental enumeration), and marking those whose deletion leads to ambiguity. The set of critical clues forms a minimal sufficient subset. Conversely, to find the maximum number of persons that can be placed under a set of negative constraints, the solver treats it as a maximum independent set problem on a conflict graph where vertices are persons and edges represent mutual exclusion (e.g., “A and B cannot be adjacent”). For small graphs (≤ 8 vertices) a brute‑force backtracking with pruning by degree ordering solves it instantly. The examiner expects the candidate to recognise the pattern and apply the appropriate graph‑theoretic insight rather than enumerate all permutations.

Another optimisation twist is the “least‑constrained first” heuristic: when multiple arrangements satisfy all clues, the puzzle may ask “Which person can occupy the most different seats across all valid arrangements?”. The solver builds the solution space by systematic branching on the most ambiguous variable, records the seat set for each person, and finally counts the cardinalities. Because the branching factor is limited by the number of negative clues, the search tree remains shallow. Mastery of this technique turns a seemingly open‑ended question into a deterministic counting exercise.

Algorithmic Approach and Complexity Analysis

Understanding the computational complexity of puzzle families helps the aspirant choose the right solving strategy under time pressure. Pure linear arrangement with only relative offsets is equivalent to finding a topological ordering of a DAG and can be solved in O(V + E) time using Kahn’s algorithm, where V is the number of persons and E the number of precedence clues. Circular arrangement adds a cyclic symmetry that can be broken by fixing an arbitrary person, reducing it to a linear problem with an extra “wrap‑around” constraint; the complexity remains linear. Double‑row and multi‑row puzzles introduce a bipartite matching component when clues link specific seats across rows (e.g., “A faces B”). This becomes a bipartite graph matching problem solvable in O(V · E) with the Hopcroft‑Karp algorithm, but because V ≤ 12 in examinations, a simple backtracking with forward checking is faster in practice. Comparison puzzles that involve only “before/after” and “immediately before/after” are also DAG topological sorts; the immediate‑precedence clues add edges of weight 1 that can be handled by collapsing the two vertices into a single super‑node before sorting.

When conditional clues, arithmetic constraints, or optimisation objectives are added, the problem becomes NP‑complete in the general case (it subsumes 3‑SAT). However, the exam instances are deliberately bounded: the number of entities never exceeds 10‑12, and the clue density is high enough to make the constraint graph almost tree‑like. The practical algorithm is therefore a depth‑first search with constraint propagation (maintaining arc consistency) and a good variable ordering heuristic — choose the variable with the smallest remaining domain (MRV). This solves any exam‑level puzzle within seconds on paper. The aspirant should internalise the MRV heuristic: always place the person or attribute with the fewest possible positions next. This mirrors the “most constrained first” principle taught in artificial intelligence courses and guarantees the smallest search tree. Recognising the underlying complexity class also prevents wasted effort: if a puzzle feels like it requires exhaustive enumeration of 10! possibilities, the solver has likely missed a structural simplification such as a symmetry break or a hidden equivalence class.

  • Treating a conditional clue as always active instead of tracking its trigger leads to false eliminations.
  • Ignoring the interaction between arithmetic constraints and the permutation of distinct values causes missed deductions.
  • Assuming that a relative clue (“A is left of B”) fixes the distance between them when it only fixes order.
  • Failing to break circular symmetry by fixing an arbitrary reference point before solving.
  • Overlooking that a negative clue (“A not adjacent to B”) also eliminates the mirrored arrangement in a double‑row facing setup.
  • Applying the MRV heuristic without updating domains after each assignment, resulting in stale candidate sets.
  • Assuming uniqueness of solution without verifying that all clues have been used; an unused clue often signals an alternative valid arrangement.

Symmetry Breaking and Reference Anchoring in Circular and Multi‑Row Puzzles

Circular arrangements possess an inherent rotational symmetry: any valid configuration can be rotated to produce another configuration that satisfies all relative clues. In examinations this symmetry is deliberately exploited to reduce the search space. The standard technique is to fix one entity — often the person with the most clues — at an arbitrary position (for example, seat 1) and treat the circle as a linear sequence that wraps around. Once the reference point is fixed, every “to the left of” or “to the right of” clue becomes a definite offset, and the wrap‑around condition is handled by a single additional constraint linking the last and first positions.

When multiple concentric rows are present (for instance, an inner circle facing outward and an outer circle facing inward), each row has its own rotational degree of freedom. The solver must anchor each row independently, but cross‑row clues such as “A faces B” couple the two reference frames. A practical approach is to anchor the inner row first, then express the outer row positions as a function of the inner row anchor plus a constant offset derived from the facing pairs. If the puzzle supplies at least one facing pair, the offset becomes known and the outer row loses its independent symmetry. If no facing pair is given, the outer row retains a free rotation; the solver must then enumerate the two possible relative rotations (clockwise or counter‑clockwise) and test each against the remaining clues.

In double‑row linear puzzles where the rows face each other, the symmetry is a reflection across the central axis. Fixing a person in the first row at the leftmost seat eliminates the reflection ambiguity. However, if the puzzle also contains a clue like “the person at the extreme left of row 1 faces the person at the extreme right of row 2”, the reflection is already broken by the clue itself, and any arbitrary anchor works. Recognising whether a clue breaks symmetry or merely restricts relative order prevents wasted branching.

A subtle trap occurs when a puzzle states “A sits opposite B” in a circular table with an even number of seats. “Opposite” fixes a distance of half the table, which simultaneously determines the rotational offset between the two persons. If the solver anchors A arbitrarily, B’s seat is forced, and the remaining symmetry is reduced to a possible 180° flip of the whole configuration. The flip is only valid when all other clues are symmetric with respect to that flip; otherwise the flip is eliminated automatically during constraint propagation.

Mastery of symmetry breaking reduces the effective branching factor from factorial to linear in the number of entities, turning a seemingly exponential puzzle into a deterministic sequence of placements.

Arc Consistency and Domain Reduction in Multi‑Attribute Matrices

Multi‑attribute puzzles assign each entity a tuple of values (for example, person → {colour, profession, floor}). The natural representation is a matrix where rows are entities and columns are attributes. Each cell initially contains the full domain of possible values for that attribute. Clues act as constraints that prune domains: a direct clue “the engineer lives on the third floor” removes all floors except three from the engineer’s floor domain and removes engineer from all other persons’ profession domains.

Arc consistency (AC‑3) systematically enforces binary constraints between pairs of variables. For every constraint “X and Y cannot share the same colour”, the algorithm revises the domain of X by deleting any colour that has no support in Y’s current domain, and vice‑versa. The process repeats until a fixed point is reached. In examination puzzles the constraint graph is sparse and the domains are tiny (usually ≤ 5 values), so a manual simulation of AC‑3 on paper is fast: list each binary clue, then iteratively cross out unsupported values until no further deletions occur.

After arc consistency, many cells become singletons. Those singletons propagate further through the remaining constraints, often solving the puzzle without any guesswork. When a puzzle still has multiple values in several cells, the solver selects the variable with the smallest remaining domain (the MRV heuristic) and tries each value in turn, re‑applying arc consistency after each trial. This depth‑first search with forward checking is the standard “backtracking with constraint propagation” method taught in artificial‑intelligence textbooks and is exactly what high‑scoring candidates execute mentally.

A common oversight is to treat a clue like “the person who likes red is not the doctor” as a simple binary inequality between two attributes of the same entity. In the matrix representation this is a unary constraint on the pair (colour, profession) for each person: the combination (red, doctor) is forbidden. Implementing it as a binary constraint between the colour column and the profession column across all persons is more cumbersome but equally correct. The solver should recognise the most compact representation to keep the domain‑reduction steps manageable.

Another nuance: some puzzles include “exactly one” constraints (e.g., “exactly one person lives on the fifth floor”). This is a global cardinality constraint that arc consistency alone cannot enforce. The practical workaround is to note the constraint, continue with binary propagation, and at the final stage verify that the cardinality holds; if it does not, backtrack to the most recent branching point.

Worked Example: Arc Consistency on a Three‑Attribute Puzzle
Four persons — P, Q, R, S — each have a distinct colour (Red, Blue, Green, Yellow), a distinct profession (Doctor, Engineer, Teacher, Lawyer), and a distinct floor (1, 2, 3, 4). Clues: (1) The Doctor lives on floor 3. (2) P is not the Engineer. (3) The person who likes Blue lives on floor 2. (4) Q likes Red. (5) The Teacher lives on floor 1. (6) S is not on floor 4. (7) The Lawyer likes Green. Determine the complete assignment.

Step 1: Initialise domains. Each person: colours {R,B,G,Y}, professions {D,E,T,L}, floors {1,2,3,4}.

Step 2: Apply unary clues. (1) Doctor → floor 3: for every person, remove floors ≠3 from Doctor profession cell. (3) Blue → floor 2: remove floors ≠2 from Blue colour cell. (4) Q colour = Red: Q’s colour domain = {Red}; remove Red from others. (5) Teacher → floor 1: similar to (1). (6) S floor ≠4: remove 4 from S floor domain. (7) Lawyer → Green: remove Green from non‑Lawyer profession cells; remove non‑Green from Lawyer colour cells.

Step 3: Enforce all‑different constraints per attribute (each colour, profession, floor used once). After Q=Red, Red removed from P,R,S colour domains. After Doctor=floor 3, floor 3 removed from non‑Doctor profession cells. After Teacher=floor 1, floor 1 removed from non‑Teacher cells.

Step 4: Iterate arc consistency. From (7) Lawyer=Green, Green removed from other professions’ colour cells. From (3) Blue=floor 2, floor 2 removed from non‑Blue colour cells. Propagate until fixed point.

Step 5: At this stage several cells are singletons: Q colour=Red; Doctor floor=3; Teacher floor=1; Lawyer colour=Green; Blue floor=2. Use all‑different to assign remaining values uniquely. The final unique solution: P=Yellow, Engineer, floor 4; Q=Red, Doctor, floor 3; R=Blue, Teacher, floor 1; S=Green, Lawyer, floor 2.

Exam‑Day Strategy
When a multi‑attribute puzzle appears, draw a compact grid with persons as rows and attributes as columns. Write the full domain in each cell using abbreviated symbols. Apply every direct clue immediately, crossing out impossible values. Then run a quick mental AC‑3 pass: for each binary clue, scan the two involved columns and delete unsupported values. Repeat until no new deletions appear. If the grid is not fully resolved, pick the cell with the fewest remaining candidates, make a provisional assignment, and continue propagation. Keep a small note of each branch so you can backtrack cleanly if a contradiction arises. This systematic approach avoids the chaotic trial‑and‑error that wastes time.

Implication Graphs for Conditional Clause Chains

Conditional clues such as “If A sits next to B, then C sits opposite D” introduce logical dependencies that are not simple binary constraints. The standard method is to translate each conditional into an implication between two literals: (A adjacent B) → (C opposite D). The negation of the antecedent (A not adjacent B) makes the implication vacuously true, so the clue only restricts arrangements where the antecedent holds. Building an implication graph where nodes are literals (positive or negative atomic statements) and directed edges represent implications allows the solver to compute the transitive closure: if X → Y and Y → Z, then X → Z. Any cycle containing a literal and its negation (X → ¬X) signals that the antecedent X cannot be true in any valid arrangement, effectively turning the conditional into a permanent prohibition.

In practice, the number of distinct atomic statements in an examination puzzle is small (typically ≤ 12). The solver can list them, draw the implication arrows, and look for forced truths. For example, if the graph yields (A adjacent B) → (C opposite D) and also (C opposite D) → (E not at extreme), then assuming A adjacent B forces a chain of deductions. If that chain leads to a contradiction with a direct clue, the assumption A adjacent B is false, so A and B cannot be adjacent. This reasoning is equivalent to unit propagation in SAT solvers and can be performed mentally with a tidy diagram.

A frequent pattern is a pair of mutually exclusive conditionals: “If X then Y” and “If X then not Y”. Their conjunction forces ¬X. Recognising this pattern instantly eliminates a whole class of arrangements without enumerating them. Another pattern is a conditional whose antecedent is a disjunction: “If A is left of B or C is right of D, then E is in the middle”. This can be split into two separate implications sharing the same consequent, which the implication graph handles naturally.

When multiple conditionals share antecedents, the graph may contain a diamond shape: X → Y, X → Z, Y → W, Z → W. The solver can deduce X → W directly, shortening the chain. The key discipline is to keep the graph updated after each definite placement; a placement may satisfy or falsify antecedents, thereby activating or deactivating whole sub‑graphs.

Bounding Techniques for Minimum and Maximum Placement

Optimization questions ask for the minimum or maximum possible number of persons satisfying a property across all valid arrangements (e.g., “What is the maximum number of persons who can sit between A and B?”). The problem reduces to finding extremal values of a linear objective over the solution set of a constraint satisfaction problem. Because the solution set is defined by a conjunction of ordering and adjacency constraints, the extremal values are attained at the boundaries of the feasible region — arrangements that push the relevant entities as far apart or as close together as the clues allow.

A practical bounding method: first, construct a “most compact” arrangement that respects all definite clues and tries to minimise the distance between the target entities. This is done by placing entities greedily from the most constrained end, always choosing the smallest available index that does not violate any clue. The resulting distance is a lower bound. Next, construct a “most spread” arrangement by placing entities as far apart as possible, typically by filling from both ends toward the centre while respecting relative order clues. The distance obtained is an upper bound. If the lower and upper bounds coincide, the answer is uniquely determined.

When the puzzle includes negative clues (“A not adjacent to B”), the compact arrangement must insert at least one separator between them, raising the lower bound. Conversely, a clue like “exactly two persons sit between C and D” fixes the distance exactly, turning the optimisation into a verification task.

For circular tables, the distance metric is the shorter arc length. The compact arrangement aligns the two entities on the same side of the anchor; the spread arrangement places them opposite each other if the table size permits. The same greedy principle applies, but the wrap‑around must be checked after each placement.

In multi‑row puzzles, the objective may involve cross‑row distances (e.g., “maximum number of persons between A in row 1 and B in row 2”). The solver treats each row independently for intra‑row bounds, then combines them using the facing or offset clues that link the rows. The overall extremum is the sum or difference of the row‑wise extrema, depending on whether the linking clues force a fixed relative shift.

Integrating Arithmetic Constraints with Ordering Logic

Hybrid puzzles add arithmetic relations such as “the sum of the floor numbers of the Doctor and the Engineer is 7” or “the age of the person who likes Red is twice the age of the person who likes Blue”. These constraints couple the ordering variables (positions, ranks) with numeric variables (floor numbers, ages). The standard approach is to treat the numeric variables as additional attributes with finite domains (usually 1‑10) and to encode each arithmetic clue as a table of allowed tuples for the involved variables. For a sum constraint X+Y=7 with domains 1‑4, the allowed pairs are (3,4), (4,3). This table is a binary constraint that can be enforced by arc consistency exactly like any other binary clue.

When the arithmetic involves ordering (e.g., “the person on the higher floor is older”), the constraint links a positional attribute (floor) with a numeric attribute (age). The solver can propagate by noting that if floor 4 is assigned to person P, then P’s age must be greater than the age of the person on floor 3. This creates a directed inequality between age variables that can be added to the implication graph as a transitive constraint.

A powerful technique is to collapse arithmetic constraints that fully determine a variable. If a clue says “the Engineer’s floor number is the square of the Teacher’s floor number” and floors are 1‑4, the only feasible pair is Teacher = 2, Engineer = 4. This instantly fixes two floor assignments and reduces the permutation space dramatically. The solver should scan all arithmetic clues for such deterministic pairs before beginning any search.

Care must be taken with “at least” or “at most” arithmetic clues (“the sum of the two highest floors is at most 7”). These are inequality constraints that prune the upper end of domains. They are handled by iteratively reducing the maximum possible value of each variable until the inequality holds for all combinations of remaining domain values. This is again a form of bound consistency, a lighter variant of arc consistency that is sufficient for linear inequalities.

Finally, when an optimisation question asks for the minimum possible sum of two numeric attributes, the solver can treat the sum as an objective and apply the bounding techniques from the previous section, using the arithmetic constraints as additional pruning rules during the greedy construction of compact and spread arrangements.

Common Traps

  • Assuming rotational symmetry is fully broken by a single anchor when a cross‑row facing clue leaves a residual 180° flip possible.
  • Neglecting to re‑apply arc consistency after each provisional assignment during backtracking, leading to stale domains and missed contradictions.
  • Treating a conditional clue as a permanent binary constraint instead of an implication that activates only when its antecedent holds.
  • Overlooking that “exactly one” global cardinality constraints cannot be enforced by pairwise arc consistency alone and require a final verification step.
  • Using the linear distance metric on a circular table without converting to the shorter arc length, producing an incorrect maximum separation.
  • Failing to recognise deterministic arithmetic pairs (e.g., square relations) that instantly fix multiple variables.
  • Ignoring the interaction between arithmetic inequalities and ordering clues, which can create hidden transitive bounds that prune domains further.

Practice Questions

← Blood Relations ← Course Home