HOW SINGAPORE EDUCATION WORKS · Probability Experiments and Theoretical Chance
Find your next learning step
Choose the question closest to your situation, or read the full guide in order.
Full chapter index · Practice and checking · Sources and next reading · How Singapore Education Works
When an experiment gives a different result from a probability calculation, check the model and the procedure before calling either one wrong. A theoretical probability follows from stated assumptions. An experimental relative frequency describes what happened in a particular set of trials. They answer related questions, but a short experiment is not required to match the prediction exactly.
This guide is for school learners and families discussing chance. It uses an original counter-drawing activity, not money or betting. The activity is a teaching suggestion rather than a prescribed Singapore assessment. Follow the student's actual syllabus and teacher's instructions for the required scope and notation.
The choices at a glance
| What you notice | Possible issue | Useful next move |
|---|---|---|
| Two colours are treated as equally likely | Category names replace elementary outcomes | Count the equally likely individual counters |
| A short run is expected to match exactly | Expected count is treated as a quota | Separate the prediction from the record |
| Unequal trial groups are averaged equally | The denominator has lost its meaning | Pool outcome counts and total trials |
| Earlier outcomes are said to make blue due | Replacement and independence are unclear | Inspect the bag before the next draw |
Full chapter index
Define the model and distinguish observations
Check the worked records and repair reasoning
Practise, recall and change the setup
Connect international mathematical learning and questions
CHAPTER 1 OF 8 · Define the model and distinguish observations
1. Establish the experiment and the event
Suppose a bag contains three blue counters and two orange counters. One counter is drawn without looking, its colour is recorded, and it is returned before the next draw. The bag is mixed again. The event of interest is drawing blue on one draw.
Assume the counters are otherwise alike and each individual counter has an equal chance of being selected. There are five possible individual counters, of which three are blue. The theoretical probability of blue is therefore 3/5, or 0.6.
There are two colour names, but that does not make their chances equal. Counting categories answers a different question from counting equally likely individual outcomes. Three blue counters represent more of the five equally likely possibilities than two orange counters.
Singapore's official 2026 Mathematics syllabus 4052 includes probability as a measure of chance and calculations from possible outcomes. That is a dated curriculum reference; use the student's own current syllabus to check examination requirements. The experimental comparison here supports understanding of the idea.
CHAPTER 2 OF 8 · Define the model and distinguish observations
2. Separate a prediction from an observed frequency
If blue occurs seven times in ten draws, its observed relative frequency is 7/10, or 0.7. That does not change the original counter count. It describes this ten-draw record under the procedure actually followed.
A theoretical probability of 0.6 does not mean blue must appear exactly six times in every ten draws. Under the model, the expected blue count for ten draws is ten multiplied by 0.6, which is six. An expected count is not a quota that every short sequence must fulfil.
Likewise, several orange results do not create a debt that the bag must repay with blue. Because the drawn counter is replaced and the starting conditions are restored, the model assigns blue probability 3/5 on the next draw too. Earlier colours do not remove counters from the bag.
Keep the assumptions visible. If the counter is not returned, the composition changes. If some counters are easier to feel or reach, equal selection may be doubtful. The probability calculation belongs to the described model; a different procedure needs a fresh account.
CHAPTER 3 OF 8 · Check the worked records and repair reasoning
3. A worked case: combine two records correctly
Assume two fictional groups use the same three-blue, two-orange bag setup with replacement. Group A records seven blue outcomes in ten draws. Group B records 22 blue outcomes in 40 draws. Both groups keep every valid trial and use the same rule for recording a colour.
| Record | Blue outcomes | Total draws | Observed blue proportion |
|---|---|---|---|
| Group A | 7 | 10 | 0.70 |
| Group B | 22 | 40 | 0.55 |
| Combined | 29 | 50 | 0.58 |
The combined calculation is (7 + 22)/(10 + 40) = 29/50 = 0.58. Simply averaging 0.70 and 0.55 gives 0.625, which assigns equal weight to groups that ran different numbers of trials. Here each draw contributes one observation, so pool the counts before calculating the proportion.
The combined observed proportion is 0.02 below the theoretical 0.60. That numerical difference alone does not prove the counters or drawing process are biased. These small fictional records illustrate calculations, not a formal statistical test.
Check the answer in another form: 29 blue and 21 orange outcomes total 50. Their proportions, 0.58 and 0.42, add to one. That confirms consistency of the arithmetic; it cannot establish that the procedure was followed accurately.
CHAPTER 4 OF 8 · Check the worked records and repair reasoning
4. Diagnose what went wrong in the explanation
If a student says the answer must be one half because there are two colours, ask them to identify the equally likely outcomes. Label the five individual counters temporarily on paper and list which three satisfy the blue event. Equal chances need a reason, not just two category names.
If they treat seven blue outcomes as a new permanent probability of 0.7, ask which statement is about the model and which is about the record. Write both sentences with their conditions: “The model gives 3/5 per draw” and “This run produced 7/10 blue.”
If they expect the next outcome to balance the last few, return to replacement. Ask what is physically in the bag immediately before the next draw. Under the stated setup it is still three blue and two orange counters, not a bag carrying a memory of previous colours.
If groups disagree about the calculation, inspect their denominators and trial rules before comparing results. Did one group omit a draw that fell on the floor? Was that rule agreed before starting? Record exceptions honestly rather than deleting awkward observations until a result looks closer to the prediction.
The broader guide to data literacy and fair comparisons develops the importance of choosing a denominator that matches the question.
CHAPTER 5 OF 8 · Practise, recall and change the setup
5. Practise a model, then inspect a record
For guided practice, assume a bag has one green counter and three yellow counters, all equally likely on a single draw. Each counter is replaced and the bag mixed between trials. State the theoretical green probability and the expected number of green results in 20 draws.
Checking answer: the probability is 1/4 and the expected count is 20 × 1/4 = 5. A fictional record of eight green results in 20 draws has observed relative frequency 8/20 = 0.4. That differs from the theoretical 0.25 without making five a compulsory observed count.
For an independent task, suppose a second run with the same setup records 12 green results in 80 draws. Combine it with the eight-in-20 record. Explain why an average of the two percentages is unsuitable for pooling these draws.
Checking answer: there are 20 green results in 100 draws, giving 0.20. The unweighted mean of 0.40 and 0.15 is 0.275, but that gives a 20-trial run the same influence as an 80-trial run. Pooling counts weights each observed draw equally.
Finally, change the rule to drawing without replacement. After a green first draw, no green counter remains in this four-counter example. Its probability on the immediate second draw is then zero. The changed answer follows from a changed setup, not from a need to balance colours.
CHAPTER 6 OF 8 · Practise, recall and change the setup
6. Return later and test a changed context
Later in the week, ask the student to explain the difference between probability and relative frequency without using the bag example. They might describe a spinner model and the proportion of trials landing in one region. Check whether they state the assumption that makes the model's outcomes equally likely.
Now consider a spinner divided into unequal areas. Counting the printed region labels alone is insufficient. The physical design and an appropriate model determine the predicted chance. Do not assume every homemade spinner works ideally just because its drawing looks balanced.
With more independent trials under a stable model, relative frequencies tend to settle near the probability. They need not move closer after every additional trial, and no finite trial count guarantees an exact match. A cumulative record can temporarily move away from the prediction while more data are added.
For a useful recall check, present a calculation with the right numerator and wrong denominator. Ask what one trial was and how many valid trials were recorded. This tests meaning rather than memory of a fraction from the earlier activity.
CHAPTER 7 OF 8 · Connect international mathematical learning and questions
7. Connect school chance with global mathematical thinking
The University of Cambridge's NRICH probability resources offer international examples that connect experiments with theoretical reasoning. They are teaching resources, not an alternative Singapore examination syllabus. The counter counts and combined-record cases in this guide are original examples.
The transferable question is whether a claim describes a model, a particular observation or an inference from observations. That distinction matters when reading an international report or a simulation result. A model can be mathematically consistent while its assumptions fit one setting better than another.
Comparing experiments across schools or countries requires more than comparing the final percentages. Check the event, materials, replacement rule, recording method and trial counts. If these differ, discuss those differences before treating the results as one combined dataset.
Continue to mathematical modelling and real-world problems when the learner is ready to ask which assumptions are useful and where a model's explanation stops.
CHAPTER 8 OF 8 · Connect international mathematical learning and questions
8. Common questions about probability experiments
Does a larger experiment always give a closer answer?
No. An additional batch can move the observed proportion farther from the prediction. More independent trials under stable conditions support long-run regularity, not a guarantee of steady improvement after each batch.
Can an experiment prove a bag is fair?
A classroom record can prompt questions about the procedure and the model. It does not prove perfect fairness. The amount and kind of evidence needed depend on the question and the statistical method being used.
Are two named outcomes always equally likely?
No. The five-counter example has two colour names but unequal colour chances. State the elementary outcomes and why they are equally likely before using a favourable-outcomes fraction.
Should an unexpected result be discarded?
Not because it is unexpected. Use a consistent rule for genuinely invalid trials, record what occurred and explain the exception. Keep the evidence separate from a wish to match the model.
Sources and next reading
Checked 22 October 2026. The official 2026 Singapore–Cambridge Mathematics syllabus 4052 supplies dated Singapore curriculum context. NRICH experimental probability resources provide the international educational connection. All numerical trial records here are fictional and are not reported experiments.
Explore scientific inquiry and evidence or return to the education hub.