Small Group Tutorials

Here to help students catch up, keep up, and move ahead. Book a consultation here.

How to Perform in the new G3 SEC Examinations | Learner’s Guide Vol 1572 | Mathematics Orchard Magnetic-Response Audit — Denominators, Probability, Graphs and Review Capacity

A percentage can be arithmetically correct yet answer the wrong question. In this G3 SEC Mathematics guide, learners use a fictional orchard magnetic-response study to practise denominators, two-way tables, simple probability, graph gradients and integer review-capacity decisions. Start with fractions and move deliberately toward mathematical modelling, while keeping the physics limitations in sight.

The 2027 Singapore-Cambridge SEC G3 Mathematics subject is K310, assessed in two written papers. Every count and operational detail below is an invented classroom example, not a real metal detector test or a validated safety procedure. EMS is the teaching series’ English, Mathematics and Science approach, not an official combined examination paper.

  • Separate 34/40 steel clips from 34/100 total objects.
  • Build a consistent two-way table and calculate simple event probabilities.
  • Interpret three distance settings without imposing an unjustified physical law.
  • Use inequalities and whole-number constraints to plan hypothetical review capacity.
  • Practise twenty-three worked questions and identify assumptions before concluding.

Vol 1571 Physics magnetic sorting · Vol 1564 Mathematics sample statistics · Vol 1568 Mathematics energy constraints · Mathematics Hub · Science Hub · Secondary English Hub · PSLE to Full SBB guide

Why arithmetic begins with a clear question

A percentage is not meaningful until we know which objects are counted and what observation qualifies as a response. The previous Physics guide used one hundred fictional classroom objects: forty labelled steel clips, thirty aluminium squares and thirty plastic tags. At two centimetres, thirty-four steel clips visibly moved within ten seconds. None of the other sixty objects met that rule. A student who writes ’34 per cent of steel clips moved’ has divided by the wrong total. This chapter turns that simple error into a wider lesson about probability, graph interpretation, integer capacity and honest conclusions. The calculations are school examples, not industrial testing results.

What the 2027 G3 Mathematics examination requires

The official 2027 SEC G3 Mathematics syllabus uses code K310, replacing the earlier 4052 reference. There are two written papers, each two hours and fifteen minutes, each worth ninety marks and fifty per cent of the overall assessment. Paper 1 contains roughly twenty-six short-answer questions; Paper 2 has about nine to ten questions, including a problem in a real-world context. Candidates attempt all questions. An approved calculator may be used in both papers. This guide practises proportional reasoning, graph reading, simple probability and inequalities. Its orchard data are original, not official paper content.

Start with the sample register

Write a register with three rows: steel forty, aluminium thirty, plastic thirty. The total is one hundred. Next add a column headed ‘visible response at 2 cm within 10 s’. Its entries are thirty-four, zero and zero. Finally add the non-response column: six, thirty and thirty. Row totals must reproduce the original sample sizes. Column totals must give thirty-four responses and sixty-six nonresponses. This two-way table prevents the learner from switching population halfway through a calculation. Before applying any percentage formula, point to the row or column whose total answers the exact question.

Two correct percentages can have different meanings

Of the forty steel clips, thirty-four responded: thirty-four divided by forty equals 0.85, or eighty-five per cent. Of all one hundred objects, thirty-four responded: thirty-four divided by one hundred equals 0.34, or thirty-four per cent. Neither number is inherently more accurate. They answer different questions. Eighty-five per cent describes a within-material response in the invented sample. Thirty-four per cent describes the response share across a deliberately mixed sample whose composition was chosen by the teacher. Neither measures the performance of a real magnetic separator.

State the denominator in a full sentence

A strong answer reads: ‘At two centimetres, thirty-four of the forty labelled steel clips met the ten-second response criterion, giving eighty-five per cent.’ A weaker answer reads: ‘The magnet was eighty-five per cent effective.’ The second wording omits the sample, response rule and meaning of effectiveness. Examination mathematics often looks short on paper, but precision in the accompanying interpretation matters. Especially in a context question, never allow a polished number to displace the actual event measured. Write the fraction first, simplify or calculate, and then write a sentence that includes its denominator.

The complete two-way table

One useful table has four headings: material, moved, did not move, total. Steel gives thirty-four, six, forty; aluminium gives zero, thirty, thirty; plastic gives zero, thirty, thirty. The grand total row reads thirty-four, sixty-six, one hundred. Each row must satisfy moved plus did not move equals total. Each column total is independently checkable. A learner can use these arithmetic checks before attempting any probability question. If the row for steel accidentally totals thirty-nine, subsequent percentages may look plausible but are built on a transcription error.

Random selection from the hundred classroom objects

If the teacher chooses one of the one hundred listed objects uniformly at random, the probability it is labelled steel is 40/100, or 0.40. The probability that it met the observed movement rule is 34/100, or 0.34. The probability it did not meet the rule is 66/100, or 0.66. These are probabilities under the stipulated random-selection model applied to the recorded sample, not observed future success rates for an external process. The selection model must be stated. Without random selection, the sample frequency is a proportion rather than necessarily a probability of the next object chosen.

Events that overlap

Let S mean that the selected classroom object is steel, and M mean that it met the movement criterion. In the stipulated table, all thirty-four responsive objects are steel, so the count for S and M together is thirty-four. Thus the probability of both events for one uniformly chosen object is 34/100. The count of steel that did not move is six, giving 6/100 for steel and non-response. Students should not add 40/100 and 34/100 to find the probability of both events. Addition addresses a different relationship and may double-count overlapping outcomes.

One event or another without double-counting

The probability of steel or a recorded movement is 40/100 in this sample, because every object that moved is already included among the forty steel clips. By the addition rule, P(S or M) equals P(S) plus P(M) minus P(S and M), or 0.40 + 0.34 − 0.34 = 0.40. The subtraction matters because the same thirty-four objects appear in both groups. For clearly disjoint categories such as steel and aluminium, add the proportions directly: 40/100 + 30/100 = 70/100. State which categories overlap before selecting an operation.

Change the information and change the denominator

Suppose the question tells us a randomly chosen object is steel. The relevant group is now the forty steel clips, not the original hundred. Under the classroom table, thirty-four of these forty moved. The appropriate fraction is 34/40, or eighty-five per cent. If instead we know the object did not move, six of those sixty-six objects are steel, so the relevant fraction is 6/66, or 1/11. These are useful extensions in reasoning about restricted groups. Do not treat a condition as decorative language; it changes the set from which we count.

Why one over eleven is easy to misunderstand

There are sixty-six nonresponsive objects in the mixed tray: six labelled steel, thirty aluminium and thirty plastic. The chance that a uniformly drawn object from this nonresponsive subgroup is steel is six divided by sixty-six, or approximately 0.0909. That is about 9.09 per cent for this invented subgroup. It is not a universal probability that a nonresponsive item contains iron, nor an assurance that nine per cent of real contaminants would be missed. The teacher deliberately chose the sample composition, and the original response rule covers visible motion rather than verified detection.

Keep the distance experiment separate

The Physics lesson also described a steel-only investigation, using forty labelled clips at each of three distances after resetting the same objects. The observed counts were thirty-four at two centimetres, twenty-one at four centimetres and eight at six centimetres. These data are not three independent groups of forty newly sampled clips. They are repeated conditions for the same set, and the lesson did not supply a series of independent replicate trials. A correct mathematical table reports the counts but does not claim the independence needed for advanced probability models or confidence statements.

Compute the three response percentages

At two centimetres, 34/40 × 100 gives 85 per cent. At four centimetres, 21/40 × 100 gives 52.5 per cent. At six centimetres, 8/40 × 100 gives 20 per cent. The fall from eighty-five to fifty-two-and-a-half is 32.5 percentage points, and the second fall is also 32.5 percentage points. It is better to report ‘percentage points’ when comparing percentages directly. Calling the decline ‘32.5 per cent’ without naming the denominator may conceal a different relative-change calculation. The arithmetic belongs to the sample, not to a law of magnetism.

Percentage points versus relative change

From eighty-five per cent to fifty-two-and-a-half per cent, the difference is 32.5 percentage points. The relative reduction compared with the initial eighty-five per cent is 32.5/85, approximately 38.2 per cent. Both calculations may appear in a mathematical argument, but they answer different questions. The first is an absolute difference in proportions; the second measures change relative to the initial figure. In context, give the base and be careful with language. A notice that merely says ‘magnetic response fell 38 per cent’ without describing the sample and conditions risks overstatement.

Use a graph whose axes name the measurements

For a distance plot, label the horizontal axis ‘separation from magnet face (cm)’ and the vertical axis ‘number of clips meeting movement rule (out of 40)’. Plot (2,34), (4,21) and (6,8). If graphing percentages instead, use y values 85, 52.5 and 20, and label the vertical axis accordingly. Counts and percentages are both valid representations if the denominator is fixed and transparent. Do not label the y-axis ‘magnetic field strength’ because no field-strength sensor produced those counts. A good axis label prevents a scientific claim from slipping into a mathematical drawing.

Why a straight line is not automatically a model

The invented points lie along a straight descending line if one joins them. Between two and four centimetres the change is minus thirteen clips over two centimetres; between four and six it is also minus thirteen clips over two centimetres. The observed segment gradient is therefore minus 6.5 clips per centimetre. That mathematical description does not establish a linear physical law. Threshold response, magnetic forces, friction and geometry do not generally vary as a simple line over all distances. Unless a question explicitly stipulates a linear model for a limited interval, do not extrapolate to eight centimetres.

A useful interpolation and its limit

If a mathematics question expressly assumes a straight-line interpolation between the measured points at two and four centimetres, the model at three centimetres gives 34 − 6.5 × 1 = 27.5 clips. A physical count cannot be half a clip. The model output is an estimated average or continuous approximation, not an observed result. A reasonable report would say the model predicts about twenty-eight responsive clips, subject to its simplifying assumption. No clip count was actually measured at three centimetres. State both the estimated value and the reason it is only a model.

Do not extrapolate a negative count

Extending the same fitted line without limits would eventually predict a negative number of responsive clips. Such a count is impossible. This contradiction exposes the domain limitation of the model. It does not mean algebra has failed; it means the mathematical relationship was extended beyond the conditions under which it was chosen. Contextual questions often reward recognition that physically meaningful counts must be whole numbers between zero and forty. When a computed model value violates a real constraint, say so rather than quietly reporting the impossible answer.

Look at nonresponses as a complementary set

Among the forty steel clips, the nonresponse counts were six at two centimetres, nineteen at four and thirty-two at six. These are obtained by subtracting each response count from forty. Their percentages are fifteen, forty-seven-and-a-half and eighty. A check is that response and nonresponse percentages sum to one hundred per cent at each fixed distance. Complement calculations are valuable because they let a learner catch arithmetic errors before interpreting a graph. But ‘nonresponsive’ should not be renamed ‘nonmagnetic’; the operational outcome does not establish material composition.

Expected counts require an explicit model

Suppose a new textbook exercise stipulates that every item in a hypothetical collection has an independent 0.85 probability of meeting the movement rule. Under that new probability model, the expected number of responses in two hundred trials is 0.85 × 200 = 170. This is not a prediction warranted by the original forty-clip classroom experiment. It is a mathematical consequence of the hypothetical assumptions supplied for the extension. Expected value does not mean precisely 170 items must respond on the next run. Real-world inference needs representativeness, repeatability and validation.

Two passes may not be independent

A tempting argument says that if a first pass has eighty-five per cent chance of response, two passes will almost certainly remove everything. If a new artificial model independently assigns a fifteen per cent miss probability on each pass, the probability of missing both is 0.15 × 0.15 = 0.0225, or 2.25 per cent. Independence is a stipulated assumption for this exercise, not a property established by the clip data. A difficult-to-move clip can remain difficult on both passes. Presenting 2.25 per cent as a tested machine failure rate would be mathematically and scientifically unsound.

Dependent outcomes need conditional care

Imagine six clips did not move in the first classroom pass and only those six are examined in a second modified arrangement. If three now move, the new observation is three of the six previously nonresponsive clips. The cumulative number that have moved at least once becomes thirty-seven of the original forty, or 92.5 per cent. But the second-pass fraction is 3/6, not 3/40, and the combined ninety-two-and-a-half per cent applies only to this invented two-stage procedure. Without a separate validation design, it does not prove a general recovery rate.

Why the sequence matters

The six clips tested a second time are selected precisely because they failed the first response rule. They are not a random fresh group of six from the original hundred. This selection can make their behaviour differ from the whole collection. The math student should track the sampling path before using probability formulas. In examinations, a tree diagram can help when branch probabilities are supplied, but the branch probabilities must correspond to the groups they describe. Do not assign eighty-five per cent to the second branch merely because it was the overall first-pass proportion.

Move from proportions to staffing capacity

The fictional orchard committee has a separate, hypothetical manual review station. Suppose its documented planning capacity is fifty-five objects per whole hour, under a specified training setup. A batch contains two hundred objects needing manual review. Dividing 200 by 55 gives about 3.636 hours at the average throughput rate. If shifts are available only as complete one-hour blocks and work must finish before release, four blocks are required. A decimal answer alone does not satisfy the whole-block scheduling condition. The correct mathematical decision uses the constraint stated by the question.

Check a two-hour deadline

A two-hour slot at fifty-five objects per hour covers at most one hundred and ten objects under the idealised constant-rate model. Two hundred exceeds this capacity by ninety objects. The inequality for one station with h available hours is 55h ≥ 200. It gives h ≥ 200/55, or approximately 3.636. If h must be a whole number, the minimum is four. For a two-hour deadline, the current one-station plan is infeasible. Stating ‘we will work faster’ is not a mathematical solution. A feasible plan must alter an explicit variable or constraint.

Find the number of review stations

If two identical independent review stations can each inspect fifty-five distinct objects per hour with no duplicated work, then two hours give 2 × 55 × 2 = 220 places. Under those stipulated conditions, two stations are sufficient for two hundred objects, with twenty spaces left. The model assumes staffing, equipment and workflow are available and that throughput scales without interference. These assumptions must be checked before converting the arithmetic into an operational promise. The calculation is a textbook exercise in whole-number planning, not certification of a real quality-control line.

Add an explicit reserve

Suppose the committee requires room for thirty additional objects beyond the planned two hundred. The capacity requirement becomes two hundred and thirty. Two stations for two hours provide only 220 and are short by ten. Three stations offer 330 under the same simplified conditions. Alternatively, a schedule with more time may work. The reserve was introduced as a new hypothetical policy parameter and must not be silently subtracted from the number of objects. In a real-world-context mathematics question, show the inequality, test feasible integer options and explain the chosen constraint.

Keep time units aligned

The review capacity is measured as objects per hour, so multiplying by hours gives a count of objects. If the available slot is ninety minutes, convert it to 1.5 hours before multiplying: one station handles 55 × 1.5 = 82.5 modelled objects. Since partially reviewed objects cannot count as completed, at most eighty-two whole objects are covered by that simple constant-rate assumption. If a question instead specifies that the fifty-five figure is a rounded average, capacity should not be treated as an exact guaranteed maximum. Units and the status of a figure both matter.

Distinguish capacity from success

A station that can review fifty-five objects per hour has a throughput measure. It has not thereby demonstrated that every reviewed object is correctly classified. Capacity calculations and classification accuracy are separate. An exam answer that says ‘the two stations process 220, therefore the system is safe’ makes an unsupported leap. Similarly, a movement result from a Physics test does not tell us how long an independent trained inspection takes. Integrated questions are solved by tracing which number answers which question, not by looking for a single impressive percentage.

When exactness matters

For probability, fractions can preserve the intended sample relationships: six over sixty-six simplifies to one over eleven. For counts of people, stations or completed objects, the final answer must be an appropriate whole number. For modelled rates, a decimal may be an intermediate step. The official syllabus generally expects non-exact numerical answers to three significant figures and angles to one decimal place unless a question specifies otherwise. But contextual rounding takes priority where fractional objects are impossible. The student must explain why a ceiling or floor is appropriate rather than applying a rote rounding rule.

A decision table makes assumptions visible

Create a table with columns for scenario, stations, hours, nominal capacity, required objects and feasible under stipulated conditions. Row A: one, two, 110, 200, no. Row B: two, two, 220, 200, yes with margin twenty. Row C: two, two, 220, 230 with reserve, no. Row D: three, two, 330, 230 with reserve, yes. A clear table supports a mathematical decision, but it does not validate the physical sorting apparatus or the training requirements. Those are different evidence questions.

A data audit is part of an examination answer

Before writing a final recommendation, check the total sample of one hundred, the forty steel-only denominator, the distinction between one condition and repeated independent observations, the units on every rate, the rounded integer staffing requirement and the scope of a probability model. These checks can be done in under a minute once a learner develops the habit. Most mistakes in complex word problems are not failures to perform multiplication. They arise when the numerator, denominator, unit or assumption belongs to a different question than the one being answered.

What to revise next

A learner who understands only fractions can begin by checking row and column totals. Once that is secure, move to event probabilities and the complement rule. Then describe graph points without inventing a scientific law. Finally solve the whole-number capacity inequalities and evaluate whether a new model’s assumptions match the original evidence. This is a staircase from foundation to advanced performance. The following English guide will not repeat these calculations; it will decide how to report them faithfully to a defined audience. The EMS chapter will then reconcile scientific limits, mathematical feasibility and public language.

Twenty-three worked examination decisions

The tasks are original examination-style reasoning exercises built from a clearly fictional dataset. Try each question without consulting the explanation first, then rewrite your answer using a named denominator, variable and unit.

Practice 1: Count the tray

Question: How many items are in the illustrative tray? Add forty steel, thirty aluminium and thirty plastic to obtain one hundred. The total must be checked before any percentage calculation. A percentage with the wrong overall count may still look tidy, which is why establishing the sample size is the first step.

Practice 2: Find the steel response share

Question: What fraction of steel clips moved at two centimetres? Thirty-four out of forty, simplifying to seventeen over twenty. In percentage form the result is eighty-five per cent. The phrase ‘of steel clips’ fixes the denominator at forty; using one hundred would answer a different question.

Practice 3: Find the all-object share

Question: What percentage of the mixed tray met the rule? Thirty-four of one hundred, or thirty-four per cent. This describes the composition and recorded responses of a particular teaching sample. It is not a universal machine detection efficiency. The answer needs a sentence that identifies the one hundred objects.

Practice 4: Complete the steel row

Question: If forty labelled steel clips were tested and thirty-four moved, how many did not move? Subtract thirty-four from forty to get six. The count six describes failure to meet the response criterion, not proof of nonmagnetic material. The complementary group must add to forty with the responses.

Practice 5: Complete the column

Question: How many mixed-tray objects did not respond? Six steel plus thirty aluminium plus thirty plastic equals sixty-six. Check 34 + 66 = 100. This column total is the denominator when a question explicitly restricts selection to the nonresponsive objects.

Practice 6: A restricted probability

Question: Given that the randomly selected classroom object is one of the sixty-six nonresponsive objects, what fraction are labelled steel? Six over sixty-six, or one over eleven. The restriction is the crucial information. The answer is about this invented table and does not generalise to all physical metals.

Practice 7: Probability of steel

Question: If one object is chosen uniformly from the full hundred, what is P(steel)? Forty divided by one hundred equals 0.4. This assumes the choice is uniform. A person deliberately selecting the heaviest objects would create a different selection process, so the probability would not necessarily remain 0.4.

Practice 8: Overlapping events

Question: In the data, how many objects are both steel and responsive? Thirty-four. Thus the joint fraction is 34/100. Do not add steel forty and moved thirty-four, because the same thirty-four responsive clips belong to both sets and would be counted twice.

Practice 9: Steel or moved

Question: Find P(steel or moved) for a uniform draw. Every responsive object in the table is steel, so the union contains forty objects. P = 40/100. The addition formula also gives 0.40 + 0.34 − 0.34 = 0.40. State the overlapping group explicitly.

Practice 10: Distance four percentage

Question: Twenty-one out of forty steel clips moved at four centimetres. Divide twenty-one by forty to obtain 0.525, or 52.5 per cent. The three distance settings must all be evaluated against forty clips, not against the one hundred mixed materials used in the separate classification demonstration.

Practice 11: Distance six percentage

Question: Eight out of forty clips moved at six centimetres. The response percentage is 8/40 = 0.20 = twenty per cent. The complementary nonresponse percentage is eighty per cent. Both numbers describe the stipulated ten-second outcome at that distance.

Practice 12: Percentage point difference

Question: What is the change from eighty-five per cent to fifty-two-and-a-half per cent? The difference is 32.5 percentage points. A relative percentage change needs a further division by the original eighty-five per cent. Naming the kind of difference prevents an ambiguous final sentence.

Practice 13: Relative change

Question: Express the first decline as a fraction of the original rate. Calculate 32.5/85 ≈ 0.38235, or about 38.2 per cent. The base of the relative change is eighty-five, not one hundred. State that the calculation refers only to the invented trial proportions.

Practice 14: Coordinate pair

Question: Give the graph coordinate for four centimetres using a count axis. The correct point is (4,21). The horizontal unit is centimetres and the vertical quantity is a count out of forty. Substituting 52.5 on a count axis would be an inconsistent graph.

Practice 15: Gradient of the segment

Question: Find the gradient between (2,34) and (4,21). Compute (21 − 34)/(4 − 2) = −13/2 = −6.5 clips per centimetre. This is an observed segment gradient, not a universal force law. Keep the unit attached to the interpretation.

Practice 16: Interpolated value

Question: Under an explicitly assumed linear interpolation from two to four centimetres, estimate the result at three centimetres. The model yields 27.5 clips. Because clips are indivisible and this is only a model, describe it as approximately twenty-eight rather than a measured half-clip outcome.

Practice 17: Invalid extrapolation

Question: Why should the line not be extended indefinitely? Eventually it predicts fewer than zero responsive clips, which is impossible. Even earlier, the shape of the actual distance response may cease to be linear. The domain of a mathematical fit depends on its evidence and constraints.

Practice 18: Independent-pass model

Question: Under a new model where misses have independent probability 0.15 on each of two passes, calculate miss-both probability. Multiply 0.15 by 0.15 to obtain 0.0225. The assumption of independence is supplied for the exercise; the original repeated-clip study did not establish it.

Practice 19: Second inspection

Question: A hypothetical retest moves three of six previously unmoved clips. How many unique original clips have moved at least once? Thirty-four plus three equals thirty-seven, giving 37/40 = 92.5 per cent. Do not add the three twice or claim it is a fresh independent sample.

Practice 20: One-station time

Question: One manual review station handles fifty-five objects per hour, and two hundred need checking. How many complete hours are needed? Two hundred divided by fifty-five exceeds three but not four. Under the stated whole-hour-block rule, reserve four hours. The ceiling follows from the scheduling condition.

Practice 21: Two stations two hours

Question: Two stations at fifty-five objects per hour for two hours provide how much modelled capacity? Calculate 2 × 55 × 2 = 220 objects. Two hundred fit with twenty places to spare under the explicit scaling assumptions. Throughput is not evidence of accurate detection.

Practice 22: Add reserve

Question: How many such stations are needed within two hours if the required count rises to two hundred and thirty? Each station provides 110 places. Two provide 220 and fail; three provide 330 and satisfy the constraint. The answer is three whole stations, assuming sufficient staffing and independent workflow.

Practice 23: Fix a claim

Question: Replace ‘the device will detect eighty-five per cent of dangerous metals’. A defensible statement is ‘In a fictional classroom test at two centimetres, thirty-four of forty labelled steel clips met the ten-second movement criterion.’ This retains the arithmetic but removes unsupported material, hazard and real-world detection claims.

Official syllabus and next steps

Confirm the official 2027 SEC G3 subject codes and read the K310 Mathematics syllabus for examination details and assessed outcomes. The following English chapter owns the public communication task, and the concluding EMS guide will reconcile Physics, Mathematics and language. Treat independent-trial probabilities in the optional examples as explicitly stipulated extensions, never as established properties of a real sorter.