Solving Mathematical Problems: A Personal Perspective — Chapter-by-Chapter Outline

Author: Terence Tao First published: 1992 (Deakin University Press, Geelong, Victoria) Edition covered: Second edition, Oxford University Press, 2006 (116 pp.). The second edition reformatted the text in LaTeX, reorganised the material into a more logical order, corrected inaccuracies, and added further exercises. The first edition (85 pp., hardcover) is out of print; its publisher no longer exists.

Central thesis

A gifted young mathematician — Tao wrote the original text at age fifteen while competing in international Mathematical Olympiads — argues that problem solving in competition mathematics is neither mystical nor purely intuitive. It is a learnable craft governed by explicit, named strategies: understand the problem before doing anything else; decompose goals into sub-goals; modify the problem slightly to build intuition before modifying it significantly; exploit every datum before introducing new ideas; and write everything down. These heuristics are then demonstrated, rather than merely stated, through worked problems in number theory, algebra and analysis, Euclidean geometry, analytic geometry, and mixed combinatorics.

The book's method is what sets it apart. Rather than presenting polished solutions, Tao recreates the experimental process — showing why one approach was tried, why it stalled, and why a different angle was chosen — so that readers learn to navigate uncertainty, not just to verify completed proofs.

How does a mathematician know where to begin when faced with a problem whose answer is unknown?

Chapter 1 — Strategies in Problem Solving

Central question

What systematic habits of mind should a solver bring to any unfamiliar mathematical problem, before attempting any computation or proof?

Main argument

Tao opens by pointing out that mathematics is unusually forgiving: unlike being physically lost, a solver can always backtrack, restart, or change the problem's frame without losing anything. This freedom is also a danger — without structure, it produces aimless wandering. The chapter equips the reader with a sequence of named moves to impose order.

Recognise the problem type. Every competition problem belongs to one of a small number of logical types: "Show that…" (prove a statement), "Evaluate…" (compute a specific value), "Find all…" (characterise a solution set), or "Is there a…?" (prove existence or provide a counterexample). Identifying the type immediately constrains which tools are relevant and what "done" looks like.

Understand the data. Before anything else, list everything the problem states. In a geometry problem this means listing the given lengths, angle relationships, and circle memberships. In a number theory problem it means listing which variables are integers, which are positive, and which divisibility constraints hold. Data that seem redundant often turn out to be the key.

Understand the objective. Clarify what the endpoint is. If the goal is to prove an inequality, one should know whether the intended path is direct, by contradiction, or by induction. If the goal is to "find all" solutions, one should note whether the answer set is expected to be finite or infinite.

Select good notation. Poor notation forces the solver to track the same object under several names and wastes mental bandwidth. Tao illustrates this with a triangle problem where labelling the sides as an arithmetic progression immediately converts the geometric constraint into a clean algebraic equation.

Write down everything you know. Documenting every observation — including apparently useless ones — serves three purposes: it provides an easy reference, it gives the solver something to stare at when stuck, and the physical act of writing triggers new connections. Tao contrasts this with the common error of "solving in one's head," which makes backtracking nearly impossible.

Modify the problem slightly. When the main problem resists attack, examine special or extreme cases (e.g., set a free parameter to 0 or to ∞), simplify a hypothesis, or formulate a related conjecture. A special case either suggests the proof structure for the general case or reveals a counterexample. Tao demonstrates with Problem 1.1: trying d = 0 in a triangle problem immediately supplies the key sub-result.

Modify the problem significantly. More drastic alterations — removing a datum, swapping the hypothesis and conclusion, generalising to higher dimensions — test which features of the problem are essential. This is how one discovers that a proof is "really" about a more general phenomenon.

Prove auxiliary results. When the main objective is opaque, prove whatever you can from the available data. Auxiliary results become lemmas that shorten the eventual proof and sometimes pivot the solver in an unexpected direction.

Exploit tactical goals. Once a sub-goal is established, focus single-mindedly on it rather than re-surveying the whole problem. Tactical goals act like waypoints: they let the solver work forward from data and backward from the conclusion until the two paths meet. Tao closes the chapter with a complete worked example — a triangle problem solved end-to-end using all the strategies introduced — demonstrating how the heuristics interact in practice, including the use of Heron's formula as the final algebraic tool.

Key ideas

• Competition problems come in distinct logical types; recognising the type is the first orienting move.
• The data in a problem are assets: listing them exhaustively before attempting a proof prevents wasted effort.
• Good notation is not cosmetic — it converts hard problems into tractable ones by exposing algebraic structure.
• Writing everything down externalises the problem and frees working memory for new ideas.
• Special cases and slight modifications build the intuition needed to attack the general case.
• "Low-risk" strategies (special cases, direct computation) should precede "high-risk" strategies (clever tricks, indirect proofs).
• A tactical sub-goal turns a vague objective into a concrete target and allows bidirectional proof search.
• No strategy should be applied mechanically; the solver must always think ahead to assess whether the approach can actually reach the goal.

Key takeaway

Problem solving is not a single act of inspiration but a disciplined sequence of moves — understand, organise, simplify, sub-goal — that can be applied methodically to any competition problem.

Chapter 2 — Examples in Number Theory

Central question

What techniques specific to the integers — modular arithmetic, divisibility, digit properties, and Diophantine constraints — are most useful at the Olympiad level, and how does one choose among them when the problem's structure is not immediately visible?

Main argument

Tao opens with a remark that captures number theory's character: it draws results from sources that are not yet fully explained. Lagrange's four-square theorem (every positive integer is a sum of four perfect squares, e.g. 30 = 4² + 3² + 2² + 1²) is a paradigm case — easy to state, non-obvious to prove, discovered by observation long before a proof was found.

Basic assertions and modular arithmetic. The chapter begins with five warm-up assertions that establish the flavour of number-theoretic reasoning: (1) for any natural number n, the numbers n and n⁵ share the same last digit; (2) a number is divisible by 9 if and only if its digit sum is divisible by 9; (3) Wilson's Theorem — (n − 1)! + 1 is divisible by n if and only if n is prime; (4) for any positive odd k, the sum 1ᵏ + 2ᵏ + … + nᵏ is divisible by n + 1; (5) there are exactly four three-digit numbers whose last two digits match those of their squares. Each assertion is proved via modular arithmetic, which the chapter treats as the fundamental language of elementary number theory.

Section 2.1 — Digits and problems. Digit-sum problems require a shift of perspective: a direct assault (treating a number digit-by-digit) is almost never the right approach. The key insight is that digit sums are related to the number's residue modulo 9. Problem 2.1 asks whether among any 18 consecutive three-digit numbers, at least one is divisible by the sum of its digits. The solution pivots to multiples of 9, observes that multiples of 18 always satisfy the condition, and then shows they must appear in any 18 consecutive integers.

Section 2.2 — Diophantine equations. A Diophantine equation asks for integer (or sometimes rational) solutions to a polynomial relation. Tao's governing principle: to "find all solutions," one must eliminate all but finitely many possibilities. Three main tools appear. Modular reduction rules out infinite families by showing the equation has no solutions modulo some small prime. Bounding exploits the fact that for large values of a variable, one side of the equation grows faster than the other. Factorisation rewrites the equation as a product of factors whose integer-factor pairs are finite in number. Problem 2.3 uses algebraic manipulation plus modular constraints to show that certain values of the variable n are impossible, eventually reducing to a finite check. Problem 2.4 demonstrates factorisation and modular arguments applied together.

Section 2.3 — Sums of powers. Problem 2.5 examines the divisibility of 1ⁿ + 2ⁿ + 3ⁿ + 4ⁿ by 5. The key is periodicity: the residues of powers repeat with a period that divides φ(5) = 4, so one need only check n modulo 4. When 4 ∤ n, the four residues cycle through {1, 2, 3, 4} modulo 5, and their sum is divisible by 5. Problem 2.6 generalises: for odd k, the sum 1ᵏ + 2ᵏ + … + nᵏ is divisible by n + 1, proved by pairing the term jᵏ with the term (n + 1 − j)ᵏ and exploiting odd symmetry. Problem 2.7 involves primes and reciprocals, ultimately deriving a divisibility condition modulo p².

Key ideas

• Number theory's power comes from modular arithmetic: residue classes turn infinite problems into finite ones.
• For digit problems, working modulo 9 (or 3) is nearly always the right reframe.
• The meta-strategy for Diophantine "find all solutions" problems is threefold: modular reduction, bounding, factorisation.
• Periodicity of residues — a consequence of Fermat's Little Theorem and Euler's theorem — is the engine behind sums-of-powers problems.
• Wilson's Theorem (p prime ⟺ (p−1)! ≡ −1 mod p) is both a characterisation of primes and a computational tool for problems involving factorials.
• A "cumbersome formula should be simplified into something more manageable" before attempting a proof.
• Intuition about what the answer "should be" must be verified rigorously; guessing the right residue class and then proving it is a standard two-step.

Key takeaway

Elementary number theory is unified by modular arithmetic: almost every technique in the chapter — digit sums, Diophantine constraints, sums of powers — reduces to reasoning about residues modulo well-chosen integers.

Chapter 3 — Examples in Algebra and Analysis

Central question

How does one extract structural information from functional equations and polynomial identities when direct computation is intractable?

Main argument

The chapter opens with a famous quotation: "One cannot escape the feeling that these mathematical formulae have an independent existence and an intelligence of their own … that we get more out of them than was originally put into them." Tao uses this to motivate a style of algebraic problem solving that listens to what the formulas say before forcing computation.

Section 3.1 — Analysis of functions and functional equations. A functional equation is a relation that a function must satisfy for all values of its argument, such as f(mn) = f(m) + f(n) for all positive integers m, n. These problems resist direct computation because the domain is infinite. Tao's toolkit has three main moves: (a) substitute special values (e.g., m = n = 1, m = n, n = 0) to derive algebraic constraints on f; (b) use induction to extend from a base case; (c) if a function satisfying the equation is conjectured to be unique, prove existence and uniqueness separately. Problem 3.1 shows that a function mapping positive integers to integers that satisfies three specific recursive inequalities must in fact be the identity function f(n) = n for all n. The proof uses induction: the inequalities force f to coincide with the identity on successively larger domains. Problem 3.2 examines a function determined by fixed values and a multiplicative property; the solution uses induction on the prime factorisation.

Section 3.2 — Polynomials. Tao lays out the foundational facts: a polynomial of degree d over a field has at most d roots (counted with multiplicity, exactly d roots in the algebraically closed case); the roots of a factor are a subset of the roots of the product; and two polynomials that agree at more points than their degree must be identical. These facts are not merely background — they are the engines of the proofs. Irreducibility (a polynomial that cannot be factored into non-trivial polynomial factors over its base ring) is introduced as a key concept; Tao gives the criterion of evaluating the polynomial at its candidate roots and examining sign changes.

Problem 3.3 involves a symmetric relation among variables and requires deriving a formula for the sum of fifth powers. The approach is to introduce elementary symmetric polynomials and exploit Newton's identities, which relate power sums to elementary symmetric polynomials, to bootstrap from known lower-order sums to the fifth-power sum. Problem 3.4 discusses irreducibility using evaluation at specific roots and sign analysis, illustrating that a polynomial with no real roots in a certain range must be irreducible over the integers.

Key ideas

• For functional equations, the first move is always substituting special values (0, 1, the argument itself, the argument's negative) to extract algebraic constraints.
• Induction on a well-ordered parameter (the natural numbers, the degree of a polynomial, the number of prime factors) is the standard technique for extending a verified base case.
• The root-counting theorem for polynomials is both a bound and a uniqueness certificate.
• Newton's identities connect power sums (p_k = x₁ᵏ + x₂ᵏ + … + xₙᵏ) to elementary symmetric polynomials; mastering them gives direct access to a wide class of symmetric sum problems.
• "There is only one or two pieces of data that can be used, so there should be a clear direction in which to go" — sparse functional equations are often easier because there are few ways to proceed.
• Move forward from data until stuck; only then move sideways (try a different substitution) or backwards (assume the conclusion and derive the hypothesis).

Key takeaway

Algebra and analysis problems at the Olympiad level are solved not by brute symbolic manipulation but by identifying the structural constraints — root counts, symmetry, inductive structure — that force the answer.

Chapter 4 — Euclidean Geometry

Central question

How does one discover the right sequence of observations — angle relationships, circle theorems, similarity and congruence — to prove a non-obvious geometric fact from a diagram?

Main argument

Tao opens the chapter with Hardy's observation that Archimedes will be remembered when Aeschylus is forgotten because mathematical ideas, unlike languages, do not die. The point is that Euclidean geometry is not an archaic exercise but a living source of transferable reasoning skills. The chapter's meta-lesson is that the beauty of geometry lies in showing non-obvious facts to be undeniably true through repeated application of obvious facts.

Thales' Theorem as paradigm. The chapter's first result is Theorem 4.1 (Thales' Theorem): the angle subtended at the circumference by a diameter of a circle is a right angle. The proof divides the inscribed triangle into two isosceles sub-triangles (each formed by a radius), uses the fact that the base angles of an isosceles triangle are equal, and concludes that the two base-angle values must sum to 90°. This proof is a model for the rest of the chapter: introduce auxiliary lines (here the second radius), convert the diagram into an algebraic statement about angles, and apply a theorem (isosceles triangle base angles) that is obvious from a figure.

Problem 4.1 — Angle bisectors in a circle. Given triangle ABC inscribed in a circle, with D, E, F the midpoints of arcs, show that AD is perpendicular to EF. The approach uses the inscribed angle theorem (an inscribed angle is half the central angle subtending the same arc) repeatedly to express all relevant angles in terms of the arc measures of ABC. This reduces the perpendicularity claim to an arithmetic identity about arc measures.

Problem 4.2 — Angle bisector lengths. Given a triangle where two specific angle-bisector lengths from the circumcentre are equal, prove that either ∠BAC = 60° or triangle BAC is isosceles. The proof uses the extended law of sines and properties of the circumcircle to translate the length condition into an angular condition, then factors the resulting trigonometric equation.

Problem 4.3 — Ratios in a rectangle. In rectangle ABCD with a specific internal construction, prove that |AB|/|FC| = |FC|/|GA| = |GA|/|AE| — a geometric progression of lengths. The approach eliminates variables one by one using the given constraints, turning the geometric proportion into an algebraic one.

Problem 4.4 — Equilateral triangle on parallel lines. Construct an equilateral triangle with one vertex on each of three given parallel lines using compass and straightedge. The key insight is to use a 60° rotation: rotating one parallel line about a point on a second parallel line by 60° produces a transformed line whose intersection with the third parallel gives the third vertex. The rotation is then reversed to find the other two vertices.

Problem 4.5 — Equal-area rectangles. Given a central rectangle R₀ surrounded by four rectangles of equal area, prove that R₀ is a square. The proof explores the area dependencies between the outer rectangles and R₀, using the constraint that the four outer areas are equal to derive that the side lengths of R₀ must be equal.

On strategy in geometry. Tao observes that the objective is often obscure at the outset — one must transform the objective (e.g., perpendicularity → angle identity → arc-measure arithmetic) before knowing which direction to pursue. Auxiliary lines, angle chasing, and coordinate transformations each work on different problem types; choosing among them requires recognising what kind of fact is to be proved.

Key ideas

• The inscribed angle theorem and properties of isosceles triangles (both arising from equal radii of a circle) underlie a large fraction of circle geometry problems.
• Auxiliary constructions — extra lines, rotations, reflections — are the equivalent of good notation in algebra: they impose structure that makes the proof transparent.
• Angle chasing works best when the objective can be translated into an identity about angle measures; coordinate geometry works best when lengths and ratios are the target.
• In constructional problems, thinking backwards (what would the final figure look like, and how can it be reached by a standard construction?) usually finds the approach.
• "When the objective is obscure, we may have to transform the objective before we know which directions to try."
• Geometry is "full of things like this: results you can check by drawing a picture and measuring angles and lengths, but are not immediately obvious" — the gap between empirical confidence and rigorous proof is what makes geometry valuable.

Key takeaway

Euclidean geometry problems are solved by translating geometric relationships (perpendicularity, similarity, equal length) into angle identities or length ratios through a chain of standard theorems applied to a diagram enriched with auxiliary constructions.

Chapter 5 — Analytic Geometry

Central question

When a problem is stated in geometric language but resists purely synthetic methods, how can vector arithmetic, coordinates, and symmetry arguments supply a decisive algebraic foothold?

Main argument

Tao opens by arguing that "the geometrical mind is not so closely bound to geometry that it cannot be drawn aside and transferred to other departments of knowledge." The chapter shows that some competition problems are most naturally geometric in appearance but most naturally solved algebraically — and that the bridge is analytic geometry, especially vector methods and symmetry.

Problem 5.1 — Regular polygon sum of squared chord lengths. A regular polygon with n vertices is inscribed in a circle of radius 1. Calculate X, the sum of the squares of the lengths of all distinct line segments (chords and sides) formed by joining pairs of vertices. Tao begins empirically: for n = 2, 3, 4, 5, 6, the values of X are computed directly and the pattern X = n² emerges as a conjecture.

Vector reformulation. The vertices are represented as unit vectors v₁, v₂, …, vₙ in the plane. The square of the distance between vᵢ and vⱼ is |vᵢ − vⱼ|² = 2 − 2vᵢ · vⱼ. Summing over all pairs: X = Σᵢ<ⱼ |vᵢ − vⱼ|² = n(n−1) − 2Σᵢ<ⱼ vᵢ · vⱼ. Let Y = |v₁ + v₂ + … + vₙ|². Expanding, Y = n + 2Σᵢ<ⱼ vᵢ · vⱼ, so X = n(n−1) − (Y − n) = n² − Y.

Symmetry closes the argument. By rotational symmetry of the regular polygon, rotating all vertices by 360°/n maps the polygon to itself, so it maps the vector sum to itself. A vector that equals its own image under a rotation by 360°/n ≠ 0° must be the zero vector. Therefore Y = 0, and X = n².

Problem 5.2 — Integer side rectangles. A large rectangle is tiled by smaller rectangles, each of which has at least one integer side. Prove the large rectangle also has at least one integer side. The proof uses an integral: define f(x, y) = sin(2πx)sin(2πy). The integral of f over any rectangle with an integer side is 0 (because ∫₀ᵃ sin(2πx)dx = 0 for integer a). So the integral over the big rectangle equals the sum of integrals over the small tiles — which is 0. But this integral is [∫₀ᴬ sin(2πx)dx][∫₀ᴮ sin(2πy)dy], which is 0 only if A or B is an integer.

Problem 5.3 — Intersecting line segments. Two line segments on a plane are "flipped" repeatedly through their intersection point. The problem asks about the long-term behaviour of this process, using vector methods to track successive positions.

Problem 5.4 — Boy escaping from teacher. A boy stands in a square pool and a teacher walks along the pool's boundary. The teacher moves at a fixed speed k times the boy's speed. The problem asks whether the boy can escape (reach the boundary before the teacher reaches the same point). By modelling both agents as vectors and studying their angular velocities, one shows the boy can engineer an escape when k < 4 by staying near the centre while forcing the teacher to traverse an arc, then sprinting to the near edge.

Key ideas

• Representing geometric points as vectors and distances as dot products converts geometric identities into linear algebra, which is often easier to manipulate.
• The sum of equally spaced unit vectors on a circle is zero — a fact that follows from rotational symmetry and solves an entire class of polygon problems.
• "Always look for ways to exploit symmetry": symmetry reduces the number of cases and often forces a quantity (here the vector sum) to equal zero.
• The trick of integrating a periodic function (sin(2πx)) over a rectangle to detect integer sides is an example of analysis serving combinatorics — a recurring cross-domain pattern.
• "Equations are solid mathematics. Not as inspirational as pictures and ideas, but the easiest to manipulate."
• Analytic geometry is most powerful when the objective is a numerical quantity (a length, an area, a sum) that can be expressed as a formula and then simplified.

Key takeaway

When geometric intuition stalls, translating the problem into vectors or coordinates and then exploiting algebraic or analytic tools — symmetry, integrals, dot products — frequently provides a clean and general solution.

Chapter 6 — Sundry Examples

Central question

How should one approach mathematical problems that do not fit neatly into any standard branch — problems involving combinatorics, game theory, graph theory, or mixed techniques?

Main argument

Tao opens with the observation that mathematics resists strict categorisation: "it is sometimes thought of as a great entity, like a tree, branching off into several large chunks … but in practice, mathematics is not this neat." The chapter gathers problems from outside the four main domains, demonstrating that the general strategies of Chapter 1 apply even when no domain-specific toolkit is available. The solver must still guess an intermediate result, prove it, and use it to reach the conclusion; the art lies in choosing the right invariant or representation.

Problem 6.1 — Chameleon colour change. On an island there are 13 grey, 15 brown, and 17 crimson chameleons. When two chameleons of different colours meet, both change to the third colour. The question: can all chameleons eventually become the same colour?

The approach is to find an invariant — a quantity that is preserved (or changes in a controlled way) under each meeting. Representing the population as a vector (g, b, c), one meeting changes (g, b, c) to (g − 1, b − 1, c + 2) or a permutation. Examining these vectors modulo 3, one notices that each meeting leaves the differences g − b, b − c, c − g unchanged modulo 3. Initially g − b = −2 ≡ 1, b − c = −2 ≡ 1, c − g = 4 ≡ 1 (mod 3). For all chameleons to become the same colour, one of g = 45, b = 0, c = 0 or its permutations must be reached. In each case, two of the differences are ±45 ≡ 0 (mod 3), but the initial differences are all 1 (mod 3). Since the invariant cannot change, uniform colour is impossible.

Problem 6.2 — Exam scores puzzle. Alice, Betty, and Carol each sit the same set of exams, and their total scores are known but their individual per-exam scores are not. By reasoning about the relationships between their totals and the constraints on individual scores, one can determine who scored what. The problem illustrates logical deduction and the strategy of systematically eliminating possibilities.

Game-theoretic examples. The chapter also includes problems about combinatorial games (determining winning and losing positions) and graph-theoretic puzzles. For winning-position problems, Tao's advice is canonical: determine all terminal positions first (positions from which no move is possible), label them "losing"; label any position from which one can move to a losing position as "winning"; iterate. The correct strategy is always "move your opponent into a losing position." One exercise involves a game on a square where the correct strategy is "always leave the opponent with a square."

Exercise 6.1 — Concert scheduling. Six musicians must each listen to every other musician. Determine the minimum number of concerts required. This is a covering problem on the complete graph K₆: each concert is a partition of the six musicians into listeners and performers, and one seeks a minimum covering. The solution uses systematic enumeration and a counting argument.

Exercise 6.2 — Grasshoppers on a line. Three grasshoppers sit on a line; at each second, one grasshopper hops over another to a symmetric point. Prove they cannot return to their original configuration after 1985 seconds. The proof uses a monovariant — a quantity that changes monotonically with each hop (in this case, the sum of pairwise distances increases or stays constant, and after an odd number of steps the configuration cannot be restored).

Exercise 6.3 — Checkerboard rearrangement. Four pieces of a 1×1 checkerboard square must be rearranged (by allowed jumps) to form a 2×2 square. The problem asks whether this is possible. The solution uses a colouring invariant: assign colours to the board positions and check that the required rearrangement would need to violate an invariant.

Key ideas

• When a problem has no obvious algebraic or geometric structure, search for an invariant: a quantity preserved by the allowed operations. If the invariant in the target state differs from the invariant in the initial state, the target is unreachable.
• Modular arithmetic provides the most common invariants in combinatorial problems (as in the chameleons problem, where residues mod 3 are preserved).
• For combinatorial game problems, determine winning and losing positions from the terminal states outward, then always move the opponent to a losing position.
• "To solve a problem in mathematics, you usually have to guess some intermediate result, which implies the conclusion but is not logically equivalent to it." In sundry problems, the intermediate result is typically the invariant or the monovariant.
• Cross-domain problems are often easier than they look because the domain restriction on technique is lifted — vectors, modular arithmetic, integrals, and graph arguments are all available.
• "Do not give up hope before any attempt has been made to solve it" — problems with very little data (like Problem 6.2) can still be resolved by finding the right structural property.

Key takeaway

Problems outside the standard branches of competition mathematics are solved by the same meta-strategy as all others: find the right representation (usually a vector, a residue class, or a graph), identify an invariant or monovariant, and show the target state violates or satisfies it.

The book's overall argument

1. Chapter 1 (Strategies in Problem Solving) — establishes the general problem-solving framework: recognise the problem type, exhaust the data, set sub-goals, use low-risk strategies before high-risk ones, and document every step. This framework is the through-line applied in every subsequent chapter.
2. Chapter 2 (Examples in Number Theory) — applies the framework to integer problems, showing that modular arithmetic is the universal language: digit sums, Diophantine equations, and sums of powers all reduce to reasoning about residue classes.
3. Chapter 3 (Examples in Algebra and Analysis) — extends the framework to functional equations and polynomial identities, where the key tools are special-value substitution, induction, and the root-counting theorem for polynomials, demonstrating that "listening to what the formulas say" is a valid and productive strategy.
4. Chapter 4 (Euclidean Geometry) — shows that geometric problems are solved by translating visual relationships into angle identities or length ratios through auxiliary constructions and standard circle/triangle theorems, not by brute-force coordinate calculation.
5. Chapter 5 (Analytic Geometry) — demonstrates that when synthetic geometry stalls, vectors and coordinates bridge the gap; symmetry arguments (the sum of equally-spaced unit vectors is zero) and analytic tricks (integrating a periodic function to detect integer sides) extend the solver's reach beyond what diagram-drawing alone can achieve.
6. Chapter 6 (Sundry Examples) — completes the argument by showing that the strategies of Chapter 1, applied with the domain-specific tools of Chapters 2–5, suffice even for problems from combinatorics, game theory, and mixed domains — the solver's invariant-hunting instinct carries over intact.

Common misunderstandings

Misunderstanding: The book teaches tricks and shortcuts, not real mathematics.

The book's goals are explicitly pedagogical and problem-solving oriented, but Tao insists throughout that every move must be justified. The strategies are heuristics for finding proofs, not substitutes for them. The final step in every worked problem is a rigorous argument.

Misunderstanding: Because it was written by a fifteen-year-old, the mathematics is elementary.

The problems worked in the book include International Mathematical Olympiad problems and problems from national competitions; several require non-trivial results (Wilson's Theorem, Newton's identities, the inscribed angle theorem, integration of periodic functions). The book is elementary in prerequisites (no calculus is assumed) but not in depth of reasoning.

Misunderstanding: Chapter 1's strategies are equivalent to Pólya's "How to Solve It."

The two works overlap — Tao explicitly acknowledges Pólya's influence — but Tao's contribution is to demonstrate the strategies in action across five mathematical domains, showing how they interact, where they fail, and how one pivots. Pólya's work is more theoretical and philosophical; Tao's is more operational and domain-specific.

Misunderstanding: The "personal perspective" of the title means the book is anecdotal or non-technical.

"Personal perspective" refers to Tao's decision to present problems as he encountered and worked through them, including the exploratory process and dead ends. The book is rigorous throughout; the "perspective" is in the pedagogy, not the mathematics.

Misunderstanding: The analytic geometry chapter requires calculus.

Problem 5.2 (integer-sided rectangles) uses a definite integral of a trigonometric function, which appears to require calculus. In fact the integral is elementary (∫₀ᵃ sin(2πx)dx = [−cos(2πx)/2π]₀ᵃ) and the key step is recognising that it equals zero when a is an integer. No knowledge of differential calculus is needed; only the formula for this specific integral.

Central paradox / key insight

The central paradox of the book is this: the most powerful problem-solving moves are also the simplest. "Understand the problem" sounds trivially obvious. "Write down everything you know" sounds like advice for beginners. "Look for an invariant" is a single sentence. Yet across 115 pages and thirty-plus worked problems, these are exactly the moves that resolve every difficulty — not clever tricks or domain-specific theorems memorised in advance.

Tao makes this explicit in the chapter on sundry examples:

"To solve a problem in mathematics, you usually have to guess some intermediate result, which implies the conclusion but is not logically equivalent to it."

The key insight is that mathematical problem solving is fundamentally a process of representation choice: the same problem stated in one form can be opaque and in another form trivial. The chameleon problem is unsolvable as a state-space search but immediately yields to modular arithmetic. The polygon chord-length problem is difficult by direct trigonometry but falls in two lines once vertices are represented as unit vectors. The "trick," in every case, is finding the right language — and the general strategies of Chapter 1 are a systematic method for searching for that language.

Important concepts

Modular arithmetic

Arithmetic on the integers modulo a fixed integer m: two integers a and b are congruent modulo m (written a ≡ b mod m) if m divides a − b. Modular arithmetic converts infinitely many integer cases into finitely many residue classes, making many number theory problems decidable.

Wilson's Theorem

The integer n is prime if and only if (n − 1)! ≡ −1 (mod n). Equivalently, (n − 1)! + 1 is divisible by n exactly when n is prime. The theorem characterises primality via factorials and is used in problems involving divisibility of factorials and products of consecutive integers.

Diophantine equation

A polynomial equation (or system) for which integer (or rational) solutions are sought. The standard strategy is to eliminate all but finitely many candidate solutions via modular reduction, bounding, and factorisation.

Functional equation

An equation that specifies a relationship a function f must satisfy for all (or all relevant) values of its argument, such as f(m + n) = f(m) + f(n) for all integers m, n. Solutions are found by substituting special values to generate constraints, then proving uniqueness via induction.

Newton's identities

Formulae relating the power sums pₖ = x₁ᵏ + x₂ᵏ + … + xₙᵏ to the elementary symmetric polynomials e₁, e₂, …, eₙ. For example, p₁ = e₁, p₂ = e₁p₁ − 2e₂, p₃ = e₁p₂ − e₂p₁ + 3e₃. They allow the computation of any power sum from the elementary symmetric polynomials, which are in turn the coefficients of the characteristic polynomial.

Inscribed angle theorem

An angle inscribed in a circle (vertex on the circle, sides passing through two other points on the circle) is equal to half the central angle subtending the same arc. A special case is Thales' Theorem: an angle inscribed in a semicircle is a right angle. This theorem is the engine of the majority of Olympiad circle geometry problems.

Invariant

A quantity associated with the state of a system that is preserved (or changes only in a controlled direction) under every allowed operation. Invariants are used to prove impossibility: if the target state has a different invariant value from the initial state, the target is unreachable. The chameleons' modulo-3 residue is the paradigm example in this book.

Monovariant

A quantity that changes monotonically (always increasing or always decreasing) under every allowed operation. Unlike an invariant, a monovariant does not rule out reaching a target state directly; it rules out returning to an earlier state (and thus proves that certain processes terminate or are irreversible).

Tactical goal / sub-goal

An intermediate result that, if proved, would imply or significantly advance the proof of the main objective. Setting sub-goals converts a single vague objective into a sequence of concrete, verifiable steps. In the strategies framework, forward search from data and backward search from the conclusion both aim to identify sub-goals where the two paths can meet.

Rotational symmetry argument

If a geometric or vector configuration is invariant under rotation by angle θ ≠ 0, then any vector quantity associated with it (such as the sum of vertex position vectors) must be fixed by the same rotation. The only vector fixed by a non-trivial rotation is the zero vector. This argument resolves Problem 5.1 (regular polygon chord sums) in a single step.

References and Web Links

Primary book and edition information

• Tao, Terence. Solving Mathematical Problems: A Personal Perspective. Second edition. Oxford University Press, Oxford, 2006. ISBN 9780199205608 (paperback), 9780199205615 (hardback).

  • Oxford University Press product page
  • Google Books preview (second edition)
  • Tao's own books page at UCLA
  • Tao's WordPress entry for the book

• First edition: Tao, Terence. Solving Mathematical Problems: A Personal Perspective. Deakin University Press, Geelong, Victoria, 1992. ISBN 0-7300-1365-0. Out of print; the publisher no longer exists.

  • Internet Archive scan of first edition

Background and author overview

• Terence Tao — Wikipedia
• Tao's blog "What's new" — problem-solving career advice post

Key ideas and related works

• Pólya, George. How to Solve It. Princeton University Press, 1945. The foundational reference that Tao acknowledges as a direct influence; covers general heuristics at a more philosophical level.
  • Princeton University Press page
• Engel, Arthur. Problem-Solving Strategies. Springer, 1998. A systematic catalogue of Olympiad techniques at a higher level than Tao's book.

Reviews and secondary commentary

• MAA Review by Loren Larson — review in the Notices of the AMS and MAA; highly recommended.
• Math Sci Notes blog review — practitioner's review emphasising the book's pedagogical strengths.
• Goodreads reader reviews