A child draws a counter from a cup and confidently says, “Red came out twice, so blue must come next.” That prediction feels sensible but may be mathematically unjustified. Probability is not a way of promising what will happen; it is a language for describing possible outcomes under clearly stated conditions.
This Mathematics in Bukit Timah lesson imagines a school fair with counters, coloured spinners and simple classroom challenges. Every game, bag, reward, attendance estimate and probability experiment is invented for education, not a claim about an event at any Bukit Timah school. No participation fee or real-money wagering is involved. The point is to teach fractions, sample spaces, tree diagrams, independent events, conditional probability and fair comparisons.
The immediate rule: for a finite experiment with equally likely elementary outcomes, probability = favourable outcomes ÷ total outcomes. A bag containing three blue counters and two red counters gives a first-draw red probability of 2/5, provided the draw is random and each counter is equally likely to be selected. Change the bag, the replacement rule or the selection method, and the probability may change.
Article contents
- Quick answer: probability requires a specified experiment
- A Bukit Timah classroom setting without claiming a real school event
- The sample space lists all elementary outcomes
- Worked example 1: exactly one head in two flips
- Probability zero, one and everything between
- Why equally likely must be justified, not merely assumed
- Worked example 2: a spinner with unequal sectors
- The complement rule makes some questions easier
- Worked example 3: at least one six in two rolls
- Independent events do not change one another’s probabilities
- Why “red is due” is not a mathematical rule
- Replacement changes the probability model
- Worked example 4: two reds with replacement
- Without replacement creates dependence
- Worked example 5: one red and one blue in any order
- A tree diagram represents conditions visibly
- Why multiplying and adding play different roles
- A probability tree must represent the whole sample space
- Worked example 6: the probability of two blue counters
- Conditional probability is a new question about known information
- Why conditional probability is not the same as reverse probability
- Mutually exclusive and independent are not synonyms
- Worked example 7: independent dice outcomes
- Count equally likely ordered dice pairs
- Worked example 8: why six is not as likely as twelve
- Theoretical probability is not the same as observed frequency
- Worked example 9: a counter experiment with observed results
- “Expected count” is not a guaranteed count
- Fairness should be defined before comparing two games
- Worked example 10: compare expected points without money
- Why a dramatic reward can conceal a low chance
- Bayes-style questions require knowing the sampling process
- Worked example 11: two possible bags, equally chosen
- Primary 3–4: fractions as parts of equal possibilities
- Primary 5–6: systematic listing and simple event counting
- Secondary 1–2: connect tree diagrams and dependence
- Upper-secondary extension: conditional structure and expected value
- Five diagnostic errors tutors should distinguish
- A twenty-minute home diagnostic
- A week of practice without fifty repetitive worksheets
- Practice challenge with model answers
- Frequently asked questions
- Curriculum, sources and useful next reading
Quick answer: probability requires a specified experiment
A probability such as 1/2 is not meaningful without an event and a model. It might mean a fair coin lands heads, or that one of two equally likely counters is drawn.
The first question is “What can happen?” The second is “Are the outcomes equally likely?” The third is “Which outcomes count as success?” Answer those before writing a fraction.
For a fair six-sided die, each face is equally likely and the probability of rolling an even number is 3/6 = 1/2 because 2, 4 and 6 meet the condition.
A Bukit Timah classroom setting without claiming a real school event
The imaginary lesson might occur in a classroom near Beauty World or Sixth Avenue, with students drawing coloured counters and organising toy spinners. These local references make the example familiar but do not establish that a named school hosted this activity.
An educational article should never invent school events, registration details or student results. The student needs an exact description of the experiment, not a misleading assertion that data were collected in the neighbourhood.
The method also works away from Bukit Timah. Mathematics transfers when the learner sees the same sample-space structure inside a new story.
The sample space lists all elementary outcomes
Flip a fair coin once. The sample space is {H, T}, representing heads and tails. If the outcomes are equally likely, each has probability 1/2.
Flip the same fair coin twice independently. The ordered sample space is {HH, HT, TH, TT}. There are four equally likely sequences, assuming fair independent flips.
The word ordered matters. HT and TH are distinct sequences, even though each contains one head and one tail. When a student merges them into one outcome, the final probability may be wrong.
Worked example 1: exactly one head in two flips
For two independent fair coin flips, the sequences with exactly one head are HT and TH. There are two favourable sequences out of the four equally likely sequences.
Therefore P(exactly one head) = 2/4 = 1/2. P(two heads) = 1/4 and P(at least one head) = 3/4.
A pupil who says “there are three possibilities: zero, one or two heads, so each has probability 1/3” is treating categories with different numbers of elementary sequences as equally likely. The sample space corrects the mistake.
Probability zero, one and everything between
An impossible event has probability 0. A certain event has probability 1. Probabilities of ordinary events lie between zero and one inclusive.
A fair six-sided die has probability zero of displaying a seven under the stated model. It has probability one of displaying a number from one to six.
For real-world uncertainty, probability zero and one require care, but in a defined classroom sample space they are precise. A learner should check the range before trusting a computed fraction such as 7/6.
Why equally likely must be justified, not merely assumed
If a spinner has three regions with unequal areas and is otherwise fair in landing over its area, its colour outcomes need not have equal probabilities. Counting three colours and assigning 1/3 each is not automatically correct.
For a spinner with sectors occupying 180°, 90° and 90°, the ideal probability of landing in the half-circle sector is 180/360 = 1/2. The two quarter-circle sectors each have probability 1/4.
A diagram’s angles, not merely its number of labels, determine the model. This is a practical test of how a geometrical representation affects chance.
Worked example 2: a spinner with unequal sectors
Imagine a fictional spinner divided into a blue semicircle of 180°, a red 90° sector and a green 90° sector. Assume every angle direction is equally likely.
Then P(blue) = 180/360 = 1/2, P(red) = 1/4 and P(green) = 1/4. Their sum is 1, as it should be for mutually exclusive sectors covering the whole spinner.
The probability of not blue is 1 − 1/2 = 1/2. That alternative calculation is the complement rule, which often shortens word problems.
The complement rule makes some questions easier
If A is an event, the probability it does not happen is 1 − P(A). This is valid because the event and its complement cover all possible outcomes without overlap.
For two independent fair coin flips, P(no heads) = P(TT) = 1/4. Therefore P(at least one head) = 1 − 1/4 = 3/4.
Counting all favourable sequences also works, but the complement method can be much simpler when the question says “at least one” across several trials.
Worked example 3: at least one six in two rolls
Roll a fair six-sided die independently twice. The chance of not rolling a six on one roll is 5/6. The chance of no six on either is (5/6) × (5/6) = 25/36.
Therefore P(at least one six) = 1 − 25/36 = 11/36. That includes one six and two sixes.
A student who calculates 1/6 + 1/6 = 1/3 has counted the double-six outcome twice. The complement method avoids that overlap.
Independent events do not change one another’s probabilities
Two independent fair coin flips are a standard example: the first outcome does not alter the modelled probability of heads on the second flip.
Even after ten consecutive heads, the eleventh fair independent flip still has P(heads)=1/2. The previous results are observations, not a force acting on the next flip.
The statement depends on independence and fairness. It does not apply without checking whether a process has a changing composition or another dependency.
Why “red is due” is not a mathematical rule
Suppose the classroom repeatedly replaces a counter after each draw, restoring the same bag of three blue and two red counters. The probability of red on every draw remains 2/5.
Three preceding blue draws do not make a red draw more likely under that independent replacement model. A sequence can look surprising while still being compatible with independent trials.
Explain the gambler’s fallacy in this educational context without encouraging betting: the mistaken belief that past independent outcomes must be balanced immediately by future outcomes.
Replacement changes the probability model
Start with a fictional cup holding three blue and two red counters. Draw one counter, record its colour, then put it back and mix thoroughly before drawing again.
Because the starting composition is restored, each draw has P(red)=2/5 and P(blue)=3/5. Under the assumed random draw method, the two colour results are independent.
If the counter is not replaced, the bag’s composition changes. The second-draw probabilities may then depend on what occurred first.
Worked example 4: two reds with replacement
In the 3-blue, 2-red bag, drawing a red and replacing it leaves two red among five before the second draw. Thus P(red then red) = (2/5) × (2/5) = 4/25.
A pupil may mistakenly write 2/5 × 1/4 because they automatically reduce both numerator and denominator. That expression belongs to a without-replacement model, not the stated one.
Read the procedure carefully. One short phrase—“put it back”—changes the second probability.
Without replacement creates dependence
Draw a red from the same bag but do not put it back. The bag now contains three blue and one red counter among four. Conditional on first drawing red, the chance that the second is red is 1/4.
Therefore P(two reds without replacement) = (2/5) × (1/4) = 1/10. This is different from 4/25 with replacement.
The multiplication is not merely “multiply the printed fractions”. It follows two stages whose probabilities are specified according to the condition created by the first stage.
Worked example 5: one red and one blue in any order
Using three blue and two red counters without replacement, red then blue has probability (2/5) × (3/4) = 3/10. Blue then red has probability (3/5) × (2/4) = 3/10.
The ordered paths are mutually exclusive, so P(one of each in either order) = 3/10 + 3/10 = 3/5.
Writing only one path would answer “red first, then blue”, not “one red and one blue in any order”. A learner must read whether the question describes an order or an unordered combination.
A tree diagram represents conditions visibly
Begin a tree with first branches red (2/5) and blue (3/5). From red, the remaining bag contains one red and three blue, giving second branches red (1/4) and blue (3/4).
From blue, the remaining bag contains two red and two blue, giving second branches red (2/4) and blue (2/4). Each set of branches leaving one state sums to one.
Multiplying along a path gives that sequence probability. Adding distinct paths that meet the requested event gives the overall probability. Tree diagrams make this rule visible.
Why multiplying and adding play different roles
Multiply conditional probabilities to follow one complete path through successive stages. Add probabilities when combining mutually exclusive complete paths.
For “red then blue”, multiply 2/5 × 3/4. For “red then blue or blue then red”, add the two disjoint path probabilities.
The words and and or can be helpful cues but are not infallible shortcuts. The mathematical relationships are path intersection and disjoint alternatives, not the presence of a particular English word.
A probability tree must represent the whole sample space
For the 3-blue, 2-red bag without replacement, the four ordered colour paths are RR, RB, BR and BB. Their probabilities are 1/10, 3/10, 3/10 and 3/10 respectively.
The sum is 1/10 + 3/10 + 3/10 + 3/10 = 1. This checks that all possibilities were accounted for and none were incorrectly duplicated.
If the sum exceeds one, look for overlapping paths or invalid branch probabilities. If it falls below one, a path may have been omitted.
Worked example 6: the probability of two blue counters
The chance of blue on the first draw is 3/5. Without replacement, two blue counters remain among four, giving a second-blue chance of 2/4.
Therefore P(BB) = (3/5) × (2/4) = 6/20 = 3/10. Under replacement it would be (3/5)² = 9/25, a different answer.
Ask which number changed in the second factor and why. The explanation should mention physical bag composition, not merely “because the denominator gets smaller”.
Conditional probability is a new question about known information
Suppose the first counter from our original bag is known to have been red and not replaced. What is the probability the second is blue?
The answer is 3/4 because the remaining bag contains three blue among four. We write this as P(second blue | first red) = 3/4.
That vertical bar is read as “given”. It signals that we are working within a restricted state of information. Students who answer 3/5 have ignored the information that red was removed.
Why conditional probability is not the same as reverse probability
P(second blue | first red) describes a chance after learning the first colour. P(first red | second blue) asks about the first result given the second result; it is a different question.
In this particular bag model, the two conditional probabilities can be calculated, but they must not be assumed equal just because both mention red and blue.
A safer practice is to state the known condition in words, identify the compatible paths or remaining objects, and only then calculate. It prevents confusing the direction of “given”.
Mutually exclusive and independent are not synonyms
Events A and B are mutually exclusive if they cannot occur together. On one fair die roll, “rolling a two” and “rolling a five” are mutually exclusive.
Events are independent if knowledge of one does not change the probability of the other. On two independent fair dice, “first die shows two” and “second die shows five” are independent but can occur together.
Two events with positive probability generally cannot be both mutually exclusive and independent, because independence would require a positive intersection probability.
Worked example 7: independent dice outcomes
Roll two separate fair six-sided dice independently. Let A be “the first die shows 2” and B be “the second die shows 5”.
P(A)=1/6 and P(B)=1/6. Their joint probability is 1/6 × 1/6 = 1/36. Both can happen in the same trial, as the ordered result (2,5) shows.
Now compare “the first die shows 2” and “the first die shows 5”. Those events are mutually exclusive on a single die and cannot be independent when both probabilities are positive.
Count equally likely ordered dice pairs
Two fair independent six-sided dice have 6 × 6 = 36 equally likely ordered pairs, from (1,1) to (6,6). The order records which die displayed each face.
Pairs summing to seven are (1,6), (2,5), (3,4), (4,3), (5,2) and (6,1). Six favourable pairs give probability 6/36 = 1/6.
The sum twelve appears only as (6,6), giving probability 1/36. The possible sums from two to twelve are not equally likely, even though there are eleven different sum labels.
Worked example 8: why six is not as likely as twelve
The ordered pairs summing to six are (1,5), (2,4), (3,3), (4,2) and (5,1): five outcomes. Thus P(sum 6)=5/36.
The sum twelve has only one outcome, (6,6), giving 1/36. Both sums are possible, but one has five times as many equally likely ordered pairs.
This example corrects the mistaken idea that “there are eleven sums, so each has probability 1/11”. Equal labels do not mean equal numbers of underlying outcomes.
Theoretical probability is not the same as observed frequency
Under a fair-coin model, P(heads)=1/2. In ten actual trials, the observed head fraction might be 7/10, 4/10 or another value. Ten flips do not force exactly five heads.
The theoretical probability describes the model. The relative frequency describes an observed experiment. They can differ without contradicting the model.
When recording classroom results, show both quantities in separate columns. Students should not alter the theoretical probability to match a small sample automatically.
Worked example 9: a counter experiment with observed results
A fictional student conducts forty replacement draws from a bag with two red and three blue counters. The theoretical red probability is 2/5=0.4, implying an expected count of 40 × 0.4 = 16 reds.
Suppose the imaginary experiment records 19 reds instead. The observed red fraction is 19/40=0.475. It differs from 0.4 but does not establish that the model is wrong.
To evaluate the randomisation method, one would need a broader investigation. A single short sequence cannot conclusively establish fairness or bias.
“Expected count” is not a guaranteed count
A class might predict sixteen reds in forty replacement draws from the 2-in-5 bag. That expectation is an average implied by the probability, not a promise of exactly sixteen.
Observed counts can fluctuate. A student who says “We got nineteen, so our multiplication is wrong” has confused prediction of a long-run mean with a deterministic forecast.
The same distinction appears in weather language: a probability refers to uncertainty in a defined event, not a guaranteed share of a single afternoon.
Fairness should be defined before comparing two games
At an imaginary school fair, suppose Game A asks each player to flip one fair coin and earns two classroom points for heads, zero for tails. Game B asks each player to roll one fair die and earns four points on six, zero otherwise.
Both games can be operated fairly in the sense that every player uses the same random mechanism and rules. Yet their average expected point awards differ.
A classroom should specify whether “fair” means equal treatment among participants, equal expected score across options, or another mathematically defined criterion. Vague fairness is not a number.
Worked example 10: compare expected points without money
For fictional Game A, expected points = (1/2)×2 + (1/2)×0 = 1 point per play. For Game B, expected points = (1/6)×4 + (5/6)×0 = 2/3 point per play.
Game A has the higher expected point award under the stated rules. That does not guarantee a participant earns more points in a particular play.
Expected value is an upper-level enrichment idea here; it is not necessary to master it before understanding ordinary probability fractions.
Why a dramatic reward can conceal a low chance
One imaginary classroom activity awards one sticker with probability 1/2. Another awards three stickers with probability 1/6. The second offers a larger possible reward, but it is less frequent.
Expected stickers are 1/2 for each: first (1/2)×1 = 1/2; second (1/6)×3 = 1/2. The expected totals are equal, though outcome variability differs.
A complete comparison explains what is being held equal. The bigger maximum reward does not automatically imply better expected results, and equal expectation does not mean the distributions are identical.
Bayes-style questions require knowing the sampling process
Suppose a teacher selects a counter from one of two bags, but does not explain how the bag is chosen. Even if the colour contents are stated, a student cannot compute the unconditional probability of red without a model of bag selection.
If the two bags are chosen with equal probability, use those weights. If one bag is chosen three times as often, the weighted probability changes.
The broader principle is that conditional data do not automatically establish the probability of the conditioning event. State the procedure fully before computing.
Worked example 11: two possible bags, equally chosen
A fictional experiment chooses Bag X or Bag Y with equal probability. X contains three red and one blue counter, while Y contains one red and three blue counters. After choosing a bag, one counter is drawn uniformly.
P(red | X)=3/4 and P(red | Y)=1/4. With P(X)=P(Y)=1/2, overall P(red)= (1/2)(3/4)+(1/2)(1/4)=1/2.
The result does not mean each bag contains half red. The overall half arises from weighting two different conditional probabilities.
Primary 3–4: fractions as parts of equal possibilities
Begin with one fair coin or a cup holding one red and one blue counter. Ask pupils to identify all possible outcomes and whether both are equally likely.
Use number lines from zero to one and picture fractions. Then change the cup to two red counters and one blue to show how an event can have probability 2/3.
Do not introduce advanced symbolic notation before learners can explain what the numerator and denominator count.
Primary 5–6: systematic listing and simple event counting
Older Primary pupils can list two-coin outcomes, compare spinner sectors, use the complement rule for simple events, and distinguish observed frequency from predicted chance as appropriate to their syllabus.
Encourage clear representations: tables, grids, simple trees and verbal explanations. Students should check whether the event they count is one ordered sequence or several alternatives.
The important progress marker is choosing an appropriate sample space rather than guessing a fraction from keywords.
Secondary 1–2: connect tree diagrams and dependence
Secondary learners can multiply along paths and add mutually exclusive paths, distinguish with-replacement and without-replacement models, and test independence through conditions.
Give two versions of the same coloured-counter problem differing in a single phrase: “replace after the first draw” versus “do not replace”. The student must change the second-stage probabilities accordingly.
A correct pair of solutions with a clear explanation demonstrates transfer. Memorising “always multiply the fractions” does not.
Upper-secondary extension: conditional structure and expected value
More advanced learners can define events symbolically, calculate conditional probability, build two-stage sampling models and examine simple expected point outcomes.
Not every extension appears in every Singapore school course. Use the relevant level’s syllabus rather than treating one article as a list of guaranteed assessed questions.
The deeper skill is to recognise which information controls each probability. A number may be exact under a complete model but unsupported when the sampling procedure is not specified.
Five diagnostic errors tutors should distinguish
One student may omit a sample-space outcome. Another may assume different categories are equally likely. A third may forget the replacement condition; a fourth may double-count “or” paths; a fifth may think previous independent results force the next outcome.
These are not interchangeable “careless mistakes”. Each requires a different targeted example and a fresh test of understanding.
A small-group tutor can ask the student to narrate the sample space, explain the selection method, and show where each fraction came from. The explanation reveals the earliest invalid decision.
A twenty-minute home diagnostic
Minutes 1–5: Write the four ordered outcomes of two independent fair coin flips. Ask for P(exactly one head)=1/2.
Minutes 6–10: From a three-blue, two-red bag, find P(red)=2/5 and P(not red)=3/5.
Minutes 11–15: Calculate P(two reds with replacement)=4/25, then without replacement=1/10. Ask why the second factor changes.
Minutes 16–20: Ask whether four previous blue replacement draws make red more likely. The learner should say no, assuming the independent random model still applies.
A week of practice without fifty repetitive worksheets
Day one is listing simple sample spaces; day two is fractions and complementary events. Day three uses unequal spinner sectors; day four covers two fair independent trials.
Day five introduces with-replacement versus without-replacement counter draws. Day six asks the learner to draw and label a tree. Day seven gives a new story where the student must choose the method without prompting.
If the child struggles with fractions, repair that prerequisite first. More probability questions alone will not solve an underlying issue with part-whole reasoning.
Practice challenge with model answers
A. A fair coin is flipped three times. Probability of all heads? Answer: (1/2)³ = 1/8.
B. Two independent fair dice are rolled. Probability of sum seven? Answer: 6/36 = 1/6.
C. A bag has one red and three blue counters. Draw without replacement. Probability of two blues? Answer: (3/4)×(2/3)=1/2.
D. From the same bag, draw with replacement. Probability of two blues? Answer: (3/4)²=9/16.
E. A fair spinner has one 180° red sector and two 90° non-red sectors. Probability of red? Answer: 1/2, not 1/3.
Frequently asked questions
Does a probability of one-half mean an event happens every second attempt?
No. It describes a modelled probability on each applicable trial, not an alternating timetable or a guarantee about a short sequence.
Are past outcomes irrelevant to every future result?
Only for independent events under an unchanged model. Without replacement or under changing conditions, earlier results can affect later probabilities.
Why are HT and TH different outcomes?
They record the order of two flips. They are distinct equally likely ordered sequences even though both contain one head and one tail.
Do I always add probabilities when a question says “or”?
No. Straight addition works for mutually exclusive alternatives. If events overlap, the overlap must be accounted for.
When is multiplication appropriate in a probability tree?
Multiply the conditional probabilities along a complete path. Add the probabilities of disjoint paths when combining alternatives.
Are these examples actual Bukit Timah school games?
No. All bags, counters, spinners and point awards are fictional classroom examples designed to illustrate mathematics.
Is expected value suitable for every Primary learner?
Not necessarily. Younger learners should first understand possible outcomes, fractions and simple probabilities; expected value can be explored when readiness and syllabus permit.
How does a parent know a child is improving?
Change only the replacement rule or redefine the event. A learner who rebuilds the sample space and explains the new probability is reasoning independently.
Curriculum, sources and useful next reading
For curriculum context, use the MOE Primary Mathematics syllabus and SEAB’s 2027 G3 Mathematics listing, which distinguishes required school content from optional enrichment. The counters and games here are invented and are not SEAB questions.
Explore Mathematics in Bukit Timah: misleading graphs for another way numbers can mislead, and Mathematics in Bukit Timah: counting trees using sampling for a distinct treatment of how observations are collected. The immutable Clementi Secondary 1 small-group Mathematics tutorial anchors the progression from worked case to independent explanation.
The idea to remember is define the experiment, identify the full sample space, check equality and dependence, then match the calculation to the actual event asked about. That makes probability a reasoning skill rather than a guess based on recent results.
Continue this Mathematics in Bukit Timah batch: Similar triangles and shadow lengths · Arithmetic and geometric sequences · Mean, median, mode and outliers. These four companion articles use distinct worked problems to build transferable Mathematics reasoning.