Did you know? A chocolate bar with 24 squares takes exactly 23 breaks to separate into individual squares, as long as each break divides one existing piece into two. It does not matter which pieces you break first. Each break increases the total number of pieces by exactly one, so there is an invisible rule controlling every possible strategy.
That is the spirit of invariants and simpler problems. For a family looking for a JC2 H3 Mathematics tutor in Bukit Timah, the most important question is not always whether the student can calculate quickly. It may be whether they can try a small version of an unfamiliar problem, find what does not change, and turn the observation into a rigorous argument.
Quick answer: An invariant is a property that remains unchanged under permitted operations. A monovariant is a quantity that consistently moves in one direction, such as increasing by one with every step. Solving a simpler problem means reducing a difficult task to a smaller case that reveals a rule or recurrence. In H3 Mathematics (9820), the official named heuristics include solving a simpler or similar problem and uncovering pattern and structure; invariants are a useful application of those heuristics rather than a separately named additional topic.
Jump to a useful explanation
- What makes a simpler case useful?
- The exact number of chocolate-bar breaks
- Parity as an invariant
- Why two-switch moves cannot produce odd parity
- Domino tilings and recurrences
- The Towers of Hanoi
- An arithmetic invariant
- Count short binary strings first
- Twenty-five worked problem clinics
- False patterns and necessary conditions
- Practical revision and tutorial checklist
- Frequently asked questions
- Official syllabus and related guides
Small cases help discover a theorem, but do not prove it
When a problem has an integer parameter n, examine n=1,2,3 before attacking the general case. Count the possibilities, draw a diagram or follow the permitted moves. Ask which feature survives when you enlarge the problem from n to n+1.
Small cases can reveal a recurrence, such as ‘every size-n configuration comes from a size-(n−1) configuration or a size-(n−2) configuration’. They can also expose a conservation law, such as an unchanged sum, parity or residue class. But a pattern seen in the first three cases is still only a conjecture until it is proved.
The right question is not merely ‘what sequence do these answers resemble?’ It is ‘what structural reason forces that sequence?’ A Fibonacci-like count is meaningful only after a valid partition identifies two smaller disjoint types of configurations.
Another simple-problem technique is to relax a restriction. Count all possible allocations before excluding those violating a capacity. Solve a geometry problem with a larger permissible region before checking whether equality can be attained in the original one. This yields a bound or starting point, not automatically a full answer.
The official H3 Mathematics 9820 content outline lists solving simpler or similar problems, uncovering patterns and considering cases as problem-solving heuristics. Students must eventually justify the general result through mathematical proof, even if the exploration began with drawings and trial calculations.
A useful teacher question is: ‘What changed from the smaller problem to the larger one, and what remained exactly the same?’ This directs attention to transitions rather than encouraging pupils to memorise the first few numerical outputs.
The chocolate-bar proof: one break, one new piece
Take a rectangular chocolate bar of m rows and n columns, giving mn small squares. The goal is to separate all mn squares by breaking one existing piece along a grid line at a time. Each legal break divides one piece into exactly two pieces.
Start with one piece. After the first break there are two. After the second break there are three. In general, after b legal breaks there are exactly b+1 pieces, because each move replaces one piece by two and increases the count by one.
When the bar is completely separated, there must be mn pieces. Therefore b+1=mn, giving b=mn−1. This is both a lower bound and the count produced by every complete legal strategy under the stated rules.
For a 4×6 bar, mn=24 and the required number is 23 breaks. The order does not matter. You may make long strips first or break around the edges; every step still adds exactly one piece.
The rule is fragile if the permitted action changes. If one action can break several existing pieces simultaneously, the piece count might increase by more than one in a single action. The 23-break conclusion therefore depends on the exact meaning of a legal move.
This is an invariant-style state relation: piece count minus break count remains one throughout the process. It converts an apparent optimisation question into a proof that all complete strategies require the same number of legal breaks.
Parity: a state property that permitted moves cannot change
Suppose a counter starts at zero and each move adds six or subtracts four. Both changes are even, so after every move the counter remains even. Therefore it is impossible to reach the odd number 2027 from zero by any finite sequence of the permitted moves.
That conclusion does not require listing every possible path. The parity of the current value is invariant, because adding or subtracting an even integer preserves whether a number is even or odd.
Can we reach every even target? For this particular move set, the moves +6 and −4 generate every even integer as an integer combination, with a suitable finite sequence of positive counts of each move. For example, 2 is reached by +6 then −4, while −2 is reached by +6 then two −4 moves.
This illustrates the difference between necessary and sufficient conditions. Evenness is a necessary condition for reachability. To conclude all even targets are attainable, an additional construction is required rather than simply assuming the converse.
Change the permitted moves to +3 or −6 and start at one. Every move is divisible by three, so the counter remains congruent to one modulo three. A target congruent to two modulo three is then impossible. Modulo arithmetic generalises parity invariants.
When studying a move-based problem, ask whether there is a conserved remainder, sign, sum, colouring or ordering constraint. If a proposed target violates it, you have an impossibility proof without exploring every move sequence.
Flipping two switches preserves the parity of lights that are on
Imagine ten lights, all initially off. A legal move flips the states of exactly two distinct lights. Can we reach a configuration with exactly five lights on? No. The number of illuminated lights is always even.
To justify that claim, examine the three possibilities for the chosen pair: two off lights become on, increasing the count by two; two on lights become off, decreasing by two; or one on and one off exchange states, leaving the count unchanged.
In all cases, the number of on lights changes by −2, 0 or +2. Starting at zero, its parity remains even after any finite sequence of moves. A state with five lights on is odd and so cannot be reached.
The proof is exhaustive because the two selected lights must have zero, one or two lights on before being flipped. We have handled every legal case rather than simply arguing from a few sample experiments.
A stronger challenge is whether every even-parity configuration can be reached from all-off when arbitrary pairs may be flipped. Yes: choose a pair of lights that should be on, flip it, and repeat for the remaining desired pairs. The construction works because an even-sized desired set can be partitioned into pairs.
The impossibility proof and the reachability construction answer different questions. A good advanced solution distinguishes ‘the target must have even parity’ from ‘every even target is actually obtainable’.
A simpler tiling problem reveals a Fibonacci recurrence
Let T(n) count tilings of a 2×n rectangle by 1×2 dominoes, which may be placed vertically or horizontally without overlaps or gaps. Start with T(0)=1, representing the empty tiling, and T(1)=1, the single vertical domino.
Now examine the leftmost part of a 2×n rectangle. If the first column is filled by a vertical domino, the remaining region is a 2×(n−1) rectangle, with T(n−1) tilings.
If the first column is not filled vertically, the top and bottom cells must each belong to a horizontal domino extending into the next column. Those two horizontal dominoes use the first two columns, leaving a 2×(n−2) rectangle with T(n−2) tilings.
These two cases cover every tiling and do not overlap. Hence for n≥2, T(n)=T(n−1)+T(n−2). The first values are T(0)=1, T(1)=1, T(2)=2, T(3)=3, T(4)=5 and T(5)=8.
The recurrence is Fibonacci-like, but the sequence name is not the proof. The proof is the exact partition of all tilings according to the leftmost columns. Every tiling must belong to exactly one of the two types, and removing the first block leaves a smaller problem.
Notice why the empty tiling was given count one rather than zero. It is the unique way to complete a region of width zero, and it makes the recurrence work naturally at n=2. Base-case conventions can be part of the structure, not arbitrary bookkeeping.
Towers of Hanoi: solving n disks with an (n−1)-disk problem
In the Towers of Hanoi puzzle, n differently sized disks begin stacked on one peg, with smaller disks above larger ones. A move transfers one top disk to another peg, and a larger disk may never be placed on a smaller disk. We want the minimum number of legal moves to transfer the full stack to another peg.
Try small cases. One disk requires one move. Two disks require three: move the small disk aside, move the large disk, then move the small disk onto it. Three disks require seven moves following the same pattern.
For n disks, the largest disk must move from the starting peg to the destination. Before it can move, all n−1 smaller disks must be moved to the spare peg. After the largest moves, those n−1 disks must be transferred onto it. Each smaller transfer is itself an (n−1)-disk problem.
This gives the recurrence M(n)=2M(n−1)+1, with M(1)=1. Solving produces M(n)=2ⁿ−1. The argument establishes an achievable strategy, and the need to clear the largest disk before its final move supplies the matching lower-bound reasoning.
The growth is dramatic: M(4)=15, M(5)=31 and M(10)=1023. These numbers should not distract from the recursive logic. Each bigger instance is built from two smaller instances plus one forced central move.
This classic puzzle is an illustrative extension of the H3 heuristic, not a claim that Towers of Hanoi appears as a separately named syllabus topic. It shows why trying the n=1 and n=2 cases can uncover a general procedure and a proof.
Greatest common divisor as a preserved quantity
Given positive integers a and b with a≥b, replace the pair (a,b) by (a−b,b). The greatest common divisor does not change: a positive integer divides both a and b exactly when it divides a−b and b.
This gives a simple invariant behind the Euclidean algorithm. Start with (252,198). Subtract to obtain (54,198). Continue applying subtraction or divisions until the numbers share their greatest common divisor as an easily visible value.
Why does the gcd stay unchanged? If d divides a and b then it divides a−b. Conversely, if d divides a−b and b then d divides (a−b)+b=a. Thus the sets of common divisors agree in both directions.
The arithmetic can be accelerated by replacing repeated subtraction with remainder division: gcd(a,b)=gcd(b,a mod b) for positive b. The same invariant is preserved even though a single division step may replace several subtractions.
The invariant supplies correctness, while a decreasing positive remainder supplies termination. Both are needed for a complete algorithmic argument. A procedure that preserves a property but never stops may not solve the stated finite problem.
This connects with working backwards and Euclidean back-substitution, which approaches the same family of ideas by reconstructing useful coefficients after the gcd is found.
Binary strings without adjacent ones: a hidden smaller problem
Let A(n) be the number of binary strings of length n containing no two consecutive ones. Count the shortest cases: A(0)=1 for the empty string, A(1)=2 for 0 and 1, and A(2)=3 for 00, 01 and 10.
Now classify a valid length-n string by its first symbol. If it begins with 0, the remaining n−1 positions form any valid string, giving A(n−1) possibilities. If it begins with 1, the next symbol must be 0, leaving a valid string of length n−2, giving A(n−2) possibilities.
For n≥2, these are disjoint and exhaustive cases, so A(n)=A(n−1)+A(n−2). The first values 1,2,3,5,8,13 reveal the Fibonacci pattern, but the case partition is what proves it.
Try changing the rule to forbid two consecutive zeros instead. Swapping every zero with one gives a bijection between the two families, so their counts are equal. This is an example of solving a similar problem by recognising a structure-preserving transformation.
Not every count that starts 1,2,3,5 must continue in the same way. The recurrence and base cases, not just the first few numerical values, determine the entire sequence.
Advanced counting exercises often become manageable when the first or last object dictates a small collection of permissible local patterns. The rest of the object then forms an instance of the original problem with a smaller parameter.
Twenty-five original worked problem clinics
1. Count pieces, not break locations
A 3×4 chocolate bar has twelve squares. Every legal break splits one existing piece into two, increasing the number of pieces by exactly one. Starting from one and finishing with twelve pieces takes eleven breaks. No strategy under these rules can use fewer or more breaks if every step is legal and the bar ends fully separated.
2. A smaller bar identifies the rule
Try a 1×2 bar: one break. Try 2×2: three breaks. Try 2×3: five breaks. The pattern suggests mn−1, but the proof is the invariant relationship ‘pieces = breaks + 1’. Without that explanation, three cases remain a conjecture rather than a theorem.
3. A move that preserves evenness
Start at zero and permit moves +6 and −4. Every state remains even, so 2027 is unreachable. The exact path does not matter. Because both moves change the value by an even number, induction on the number of moves proves the parity invariant for every finite sequence.
4. A reachable even target
From zero, +6 followed by −4 reaches two. Adding two more −4 moves after another +6 can reach zero again. These constructions demonstrate some even targets. A parity invariant proves odd targets impossible, but it does not alone establish that all even targets are reachable; that converse needs a separate construction.
5. A modulus-three invariant
Start at one and allow moves +3 or −6. Every move is a multiple of three. Hence each state remains congruent to one modulo three. A target such as 2027 leaves remainder two modulo three and cannot be reached, regardless of the number of moves.
6. The two-switch rule
Start with ten lights off. Every move flips exactly two lights. The number illuminated changes by −2, zero or +2, so the number on always has even parity. A target with five lights on is unreachable. The proof covers all three possible on/off patterns of the selected pair.
7. Construct an even light pattern
If exactly six designated lights should be on and all others off, partition the six into three pairs and flip each pair once. Every desired light is activated, and no other light changes. This shows every even-sized set of desired on lights is reachable when arbitrary pairs may be chosen.
8. Two piles with conserved total
Start with piles containing five and nine counters. A legal move transfers two counters from one pile to the other, provided the source has at least two. The total remains fourteen. Hence a target with total fifteen is impossible. A target with the same total is not automatically attainable, because each pile’s parity is also preserved.
9. The second pile invariant
From piles (5,9), transferring two from the second to the first produces (7,7), which is reachable. The target (8,6) has the same total fourteen, yet is impossible because both starting pile counts are odd and transferring two preserves their parity. One invariant can rule out more configurations than another.
10. Domino tilings of width zero
For 2×n domino tilings, define T(0)=1 as the one empty tiling. This is not a physical domino configuration; it is the correct counting base case. It makes the recurrence T(n)=T(n−1)+T(n−2) work naturally for n=2.
11. Two by two tilings
A 2×2 board has exactly two domino tilings: two vertical dominoes or two horizontal dominoes. Thus T(2)=2. Examining the first column separates the possibilities and suggests the valid recurrence for all larger widths.
12. Two by three tilings
Using T(3)=T(2)+T(1), with T(2)=2 and T(1)=1, gives three tilings. One begins with a vertical domino and leaves a 2×2 board; the others begin with two horizontal dominoes and leave a 2×1 board. The partition is exhaustive and disjoint.
13. Do not guess Fibonacci from numbers alone
Suppose a count begins 1,1,2,3,5. Many different rules could produce that finite prefix. To establish a Fibonacci recurrence, explain exactly why every size-n object belongs to one of two nonoverlapping families corresponding to sizes n−1 and n−2.
14. One disk in Towers of Hanoi
One disk moves directly to its destination in one legal move. This sets M(1)=1. The starting case is necessary for deriving M(n)=2M(n−1)+1. Without a base case, the recurrence by itself does not determine a unique sequence.
15. Three disks in Hanoi
To move three disks legally between pegs, first transfer the two smaller disks to the spare peg in three moves, then move the largest disk once, then transfer the smaller two onto it in three moves. The total is seven. The same breakdown generates the general construction.
16. Hanoi’s exponential formula
The recurrence M(n)=2M(n−1)+1 and M(1)=1 gives M(2)=3, M(3)=7 and M(4)=15. Guess M(n)=2ⁿ−1 and verify it inductively: if M(n−1)=2ⁿ⁻¹−1, then M(n)=2(2ⁿ⁻¹−1)+1=2ⁿ−1.
17. The gcd survives subtraction
For positive a≥b, gcd(a,b)=gcd(a−b,b). If a number divides both a and b, it divides a−b; if it divides a−b and b, it divides a. Thus the sets of common divisors coincide. The operation changes the numbers without changing their greatest common divisor.
18. An actual Euclidean step
For 252 and 198, subtracting yields 54 and 198. The gcd remains 18. More efficient division uses 252=198+54, then 198=3×54+36 and 54=36+18. The remainder decreases to zero, showing both the preserved gcd and the progress of the algorithm.
19. Why a decreasing quantity matters
An invariant alone does not prove an algorithm eventually stops. In the Euclidean algorithm the positive remainders strictly decrease until zero is reached; no infinite strictly decreasing sequence of nonnegative integers exists. A correctness proof needs both a preserved target property and a termination argument.
20. Binary strings of length two
The binary strings of length two with no adjacent ones are 00,01 and 10. The string 11 is forbidden, so A(2)=3. This small enumeration exposes the restriction and prepares the partition by a string’s first symbol.
21. Count length-three binary strings
Valid three-bit strings with no adjacent ones are 000,001,010,100 and 101, giving A(3)=5. The recurrence A(3)=A(2)+A(1)=3+2 matches the enumeration. The recurrence’s proof, not this one numerical check, establishes the general claim.
22. Flip digits to find a bijection
Map every valid binary string with no adjacent ones to a string with no adjacent zeros by swapping zero and one in every position. The transformation is its own inverse, so it is a bijection. The two collections have identical counts at every fixed length.
23. Grid paths as a smaller-count problem
A route on a grid that uses exactly two right moves and two up moves has six possible step orders, C(4,2)=6. For m right and n up moves, choose the positions of the right moves among m+n total steps. This count assumes no additional obstacles or forbidden intermediate positions.
24. A colouring invariant in walking
On an integer grid, every move one unit north, south, east or west switches the parity of x+y. Starting at (0,0), after an even number of moves the endpoint has even x+y; after an odd number it has odd x+y. Therefore two legal moves cannot end at (1,0).
25. A necessary but insufficient test
Suppose a puzzle requires two moves and the target has the same parity as the starting state. That parity match is necessary under single-unit grid moves, but it does not guarantee reachability: a target ten units away cannot be reached in two unit moves. Combine an invariant with a distance or construction argument when needed.
When a good-looking pattern is not yet a proof
The first danger is overgeneralising from a short table. The values 1,2,3,5 may look familiar, but a table alone does not establish the next value. A valid recurrence needs a complete partition of every configuration and a rule that identifies the smaller remaining problem.
The second danger is treating an invariant as an if-and-only-if condition. A reachable state must satisfy all invariants, but satisfying one parity or residue restriction does not prove that a state is reachable. The target may violate another constraint, such as nonnegative piles, capacity or the minimum number of moves.
The third danger is misdescribing the move. In the chocolate puzzle, every break must split one existing piece into exactly two. If an action can cut through many pieces simultaneously, the piece-count argument does not describe the new process. Mathematical conclusions depend on permitted operations, not merely on the initial picture.
The fourth danger is forgetting a base case. A recurrence can generate many sequences depending on its starting values. Define the empty tiling or zero-length string carefully, justify the convention and verify that the recurrence begins at the stated index.
The fifth danger is stopping after showing that an algorithm preserves a property. An algorithm may keep an invariant forever without reaching a final answer. A termination proof or an explicit finite construction is often required. In the Euclidean algorithm, decreasing nonnegative remainders provide that additional argument.
The sixth danger is counting overlapping types as though they were disjoint. For domino tilings, the first-column alternatives are mutually exclusive; for other puzzles they may not be. Check that every outcome is counted once. The inclusion–exclusion guide shows how to repair more complex overlaps.
A six-session progression for JC2 H3 preparation
Session one should focus on identifying a move-based invariant. Give the pupil a step-size puzzle and ask whether parity, a residue or a total is preserved. Require an impossibility proof, not simply an unsuccessful search through examples.
Session two uses smaller cases. Count 2×1, 2×2 and 2×3 domino tilings before attempting the general n case. Ask the student to identify an exhaustive and disjoint partition by the first column. The goal is deriving a recurrence rather than memorising Fibonacci numbers.
Session three introduces a second invariant or a necessary-versus-sufficient distinction. Use the two-pile counter game: total and parity are both preserved, so one condition alone may be too weak. Then ask for a construction that reaches a permitted target.
Session four compares two recursive settings, domino tilings and Towers of Hanoi. The first has an additive recurrence of two previous terms; the second has a doubling recurrence. Students should explain why different structural partitions produce different growth patterns.
Session five studies an algorithmic invariant through Euclidean division. Have the learner justify both gcd preservation and termination. Then show how a simple remainder relation can generate a proof rather than merely calculate an answer.
Session six combines these ideas in an unfamiliar problem. Ask the student to solve a small instance, conjecture a pattern, look for a counterexample and produce a general argument. Finish by checking whether the proposed condition is necessary, sufficient or both.
For Bukit Timah families, an effective tutor should ask ‘What stays the same?’ and ‘What becomes smaller?’ rather than immediately supplying a formula. The immutable eduKateSG three-student Mathematics reference near Sixth Avenue MRT describes a Secondary 1 teaching format, not a confirmed JC2 H3 programme. Enquire directly about current H3 coverage and availability.
A simple parent review is to ask the learner to explain why the target is impossible, not just which moves they tried. A strong explanation names the invariant and shows every legal move preserves it. For recursive counts, ask the pupil to explain why the cases cover every possibility and do not overlap.
Progress can be tracked through increasingly unfamiliar problems. First the student identifies a familiar parity invariant when told to look for one. Later, without a method hint, they recognise a preserved remainder or an exhaustive recurrence. The latter is more convincing evidence of transferable H3 problem-solving ability.
Frequently asked questions
Are simpler problems and patterns part of H3 Mathematics 9820?
Yes. The 2026 syllabus names solving a simpler or similar problem and uncovering pattern and structure as problem-solving heuristics. Invariants are a useful way to apply those techniques, not a separately labelled new syllabus topic.
What is an invariant?
It is a property preserved by every legal transformation. Parity, a total sum, a congruence class or a greatest common divisor can be invariant, depending on the permitted operations.
What is the difference between an invariant and a monovariant?
An invariant remains unchanged, while a monovariant consistently moves in one direction, such as a strictly decreasing nonnegative remainder sequence used to prove that an algorithm terminates.
Does an invariant prove every matching target is reachable?
No. It provides a necessary condition under the permitted moves. Reaching a target may require additional conditions and an explicit construction or sufficiency proof.
Why do small cases help with a recurrence?
They expose possible ways to build a larger object from smaller ones. The rigorous recurrence follows only after showing that the decomposition includes every valid object exactly once.
Why is an empty tiling sometimes counted as one?
It represents the unique choice to place no dominoes in an empty region. This convention makes counting recurrences correct at their smallest inputs.
Is Towers of Hanoi officially named in the H3 syllabus?
The puzzle is an illustrative application of recursion and solving smaller problems. It is not presented here as a separately named additional syllabus topic.
How do I distinguish proof from checking examples?
Examples can support a conjecture and reveal counterexamples. A proof must cover every permitted case through induction, an invariant argument, a bijection, exhaustive cases or another justified method.
How should a student test a proposed recurrence?
Verify starting values, check several small instances, and most importantly prove that the rule partitions all size-n objects into the stated smaller subproblems without overlap or omission.
What should a good H3 explanation include?
A precise statement of the permitted moves or configurations, a preserved property or structural partition, a proof that it applies universally, and a conclusion matching exactly what the question asks.
Official reference and related guides
The 2026 Singapore–Cambridge H3 Mathematics (9820) syllabus lists uncovering pattern and structure, solving simpler or similar problems, considering cases and working backwards among its heuristics. For subsequent cohorts use the SEAB 2027 A-Level syllabus index.
Continue with induction and the pigeonhole principle, bijections and distribution counts and mathematical investigation and conjecture testing.
The lesson is to begin small when a problem looks forbidding, but finish with a general proof. A preserved property can rule out a target at once; a sound recurrence can count an enormous collection without enumeration. Both depend on carefully stating what legal operations or configurations are allowed.