How do you get an A in H2 Mathematics? There is no honest single trick, fixed percentage promise or shortcut that guarantees an A. The useful answer is to build a system that can repeatedly convert mathematical knowledge into marks across both 9758 papers: secure the assumed algebra and calculus floor, understand the H2 concepts deeply enough to select methods in unfamiliar questions, keep probability and statistics as a full part of the course, use the graphing calculator intelligently, show enough mathematical working, and practise paper control until performance is stable under time.
This page owns the search intent how to get A in H2 Mathematics and H2 Maths A grade tips. It does not replace the H2 Mathematics 9758 topic map, the first-three-months transition guide, or the separate prelims revision plan. Its job is to define what distinction-level control looks like and how to build it without turning “get A” into a promise.
For the 2027 Singapore–Cambridge H2 Mathematics syllabus 9758, SEAB gives approximate assessment-objective weightings of 30% for mathematical techniques, 60% for problem formulation and solving, and 10% for mathematical reasoning and communication. Two three-hour papers each carry half the total assessment. Paper 1 is Pure Mathematics; Paper 2 combines Pure Mathematics with a substantial Probability and Statistics section. Questions may integrate ideas from more than one topic and may use real-world contexts. Those facts matter because an A-level system cannot be built on formula recall alone.
Grade thresholds can vary between examinations and are not something this page will invent. The student’s task is therefore to maximise controllable quality: knowledge, transfer, accuracy, communication, speed, recovery and checking. A grade is the examination outcome. The training system should focus on the capabilities that can be improved.
1. Distinction is a reliability problem before it is a difficulty problem
Many students respond to an A target by searching for harder questions. Hard questions matter, but they are not the first distinction. A high H2 Mathematics grade usually requires ordinary and medium-demand marks to be converted with very little leakage. A student who can solve spectacular problems but loses routine marks through algebra, incomplete working, rounding or calculator input has built a high ceiling on an unstable floor.
The first distinction-level question is therefore: how many reachable marks are being lost for reasons that do not require more advanced Mathematics? Those losses include copied values, sign errors, missed domain restrictions, forgotten conditions, incomplete statistical conclusions, unit mistakes, premature rounding, misread graphs, poor time allocation and failure to return to skipped questions.
Reducing this leakage often produces more reliable improvement than immediately raising question difficulty. Once the base is clean, difficult problems become a genuine extension rather than compensation for avoidable loss.
2. The A-grade system has seven layers
- Assumed floor. Algebra, functions, trigonometry, exact form and calculus prerequisites must be available without excessive search.
- Concept depth. Definitions, conditions and relationships must be understood beyond procedure.
- Method selection. The student must recognise structure when the chapter label disappears.
- Execution accuracy. Algebra, notation, graphing-calculator use and probability/statistics calculations must remain controlled.
- Transfer. Knowledge must survive changed representations, mixed topics and unfamiliar contexts.
- Paper control. The student must allocate time, leave and return, recover after difficult questions and finish with usable checking.
- Feedback loop. Errors must be converted into targeted repair and then retested until they stop returning.
The layers interact. Faster execution creates time for harder questions. Better method selection reduces wasted algebra. Stronger checking converts near-misses. Cleaner assumed knowledge gives working memory back to new H2 ideas. A distinction system is therefore an integrated operating system, not a collection of isolated “tips”.
3. First secure the assumed Additional Mathematics floor
The 2027 H2 syllabus explicitly assumes O-Level Mathematics and lists assumed O-Level Additional Mathematics knowledge. That includes quadratics, surds, polynomials and partial fractions, exponentials and logarithms, coordinate geometry, trigonometry, and differentiation and integration. If these ideas remain fragile, every H2 topic pays a hidden tax.
A student aiming for an A should not interpret this as “revise all A-Math again”. Diagnose the active dependencies. If H2 integration errors come from partial fractions, repair that. If function questions collapse because logarithmic manipulation is weak, repair that. If vectors are slowed by algebraic fractions, the vector chapter is not the first repair.
The distinction-level floor is reached when assumed knowledge is sufficiently automatic that attention can stay on the new H2 structure rather than on recovering old procedures.
4. Learn definitions and conditions like working tools
At higher levels, many errors are not calculation errors. They are condition errors. A function inverse is attempted without appropriate one-to-one restriction. A geometric-series sum to infinity is used without checking convergence. A binomial model is used without checking assumptions. A hypothesis test is framed about a sample statistic rather than the population parameter. A regression line is extrapolated far beyond the observed data.
Definitions and conditions protect method selection. They tell the student when a method is legal. One efficient A-grade habit is to attach every formula or procedure to its condition: what object is this for, what must be true, and what would make it inappropriate?
This reduces a whole class of “careless” errors that are actually boundary errors.
5. Topical mastery is necessary but not sufficient
Topical practice is where new methods become fluent. The danger is staying there too long. A worksheet titled “Integration by Parts” has already solved the first exam problem for the student: it has named the method. A mixed paper does not.
The transition from topic practice to mixed practice should happen early. Once a method is reasonably secure, place it among competing methods. Ask the student to identify the structure before calculating. Then vary the surface, representation and context. Method recognition should become part of every topic’s mastery definition.
The student is closer to distinction when a question can change appearance without changing their control.
6. Paper 2 statistics is not a side subject
Probability and Statistics carries substantial weight in Paper 2. Students who postpone it because Pure Mathematics feels more prestigious are taking a structural risk. The statistics section requires its own language: random variables, distributions, sampling, hypothesis testing, correlation and regression, modelling assumptions, calculator workflows and contextual conclusions.
A-grade preparation keeps statistics alive throughout the course. Definitions and model conditions should be retrieved regularly. Calculator routines should be practised with interpretation rather than button memory. Written conclusions should answer the context rather than merely report “reject H0”.
Pure Mathematics and Statistics also reward different habits. Pure work often emphasises transformation and structural recognition; statistics demands modelling, assumptions and interpretation. Strong students learn to switch modes deliberately.
7. Use the graphing calculator as an instrument, not an oracle
The approved graphing calculator can remove computation and visualisation load, but it can also create false certainty. A displayed graph may hide roots because of the window. A numerical solver may return one value when more exist. A regression output may be copied from the wrong orientation. Premature rounding can corrupt later work. A decimal can replace an exact form that the question still needs.
Distinction-level GC use begins with prediction. What should the graph roughly look like? How many roots are plausible? What sign and magnitude should the answer have? What variable is being predicted? Then use the calculator, and check whether the output agrees with mathematical structure.
The calculator should make a strong mathematician faster; it should not make an uncertain mathematician less visible.
8. Show working that preserves mathematical evidence
SEAB’s syllabus notes that unsupported graphing-calculator answers may not receive credit in contexts where working is required, and method marks can depend on evidence of a correct mathematical process. Good working is therefore not decorative handwriting. It preserves the logic of the solution.
A distinction script usually avoids both extremes: not every trivial arithmetic step is written, but transformations, equations, substitutions, conditions and key reasoning are visible. Another reader can reconstruct the route. This also makes checking easier because the student can inspect decisions rather than one final number.
9. The error ledger should become a strategy ledger
An ordinary error log records what went wrong. A stronger one also records the better decision. “Forgot +C” is weaker than “For every indefinite integral, check whether the answer represents a family of antiderivatives before leaving the line.” “Wrong regression line” becomes “State which variable is predicted from which variable before reading GC output.”
This turns feedback into future control. The aim is not to collect mistakes. It is to build a library of decision rules that the learner increasingly runs without external prompting.
10. Twenty distinction checkpoints across the H2 system
1. Algebra should disappear into the background
Distinction standard. At distinction level, routine manipulation should not consume the attention needed for the H2 idea itself. Algebraic fractions, factorisation, exact form and change of subject should remain reliable under mixed load. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Audit algebra inside real H2 questions rather than only in isolated drills. Record whether the student’s first wrong line is mathematical strategy or symbolic execution. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. If algebra is still the main source of leakage, repair it before buying harder H2 questions. The best sign of progress is not faster algebra alone but more working memory available for the higher-level structure. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
2. Functions must be read structurally
Distinction standard. A-grade work distinguishes domain, range, inverse, composition and transformation rather than treating function notation as decoration. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Before calculation, ask what object is being transformed, what restrictions matter, and whether an inverse exists on the stated domain. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. Use graph, algebra and mapping representations of the same function. Distinction control appears when the student can move among them without losing conditions. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
3. Graphs should be predicted before they are plotted
Distinction standard. The GC makes graphing easy, but a distinction student should have an expectation before the screen appears. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Predict intercepts, asymptotes, symmetry, end behaviour and turning behaviour before plotting. Then use the calculator as a check. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. This habit protects against bad windows, hidden features and blind trust in technology. It also improves sketching when the question demands mathematical evidence rather than a screenshot-like answer. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
4. Equations and inequalities need boundary discipline
Distinction standard. Many high-value losses come from invalid transformations, missed restrictions or incomplete solution sets. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Train the student to mark domain restrictions, sign-changing operations and excluded values before manipulating aggressively. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. For inequalities, compare algebraic and graphical routes. For equations, substitute or reason back into the original condition when extraneous solutions are possible. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
5. Sequences require object clarity
Distinction standard. Students often confuse a term, a partial sum and a limiting sum because notation is read too quickly. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Write what u_n and S_n represent in words before using formulas, especially in mixed progression questions. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. Distinction-level work treats convergence conditions and interpretation as part of the method, not as small details added after calculation. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
6. Series methods should be selected, not guessed
Distinction standard. Method of differences, geometric structure and other sequence tools can look similar on the page. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Ask what pattern makes the sum tractable before launching algebra. Verify decompositions with a few small n-values. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. The goal is to see why the method compresses the sum. A correct memorised decomposition that the student cannot validate remains fragile. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
7. Vectors should remain geometric while algebra runs
Distinction standard. Vector questions can become long symbolic systems in which the learner forgets the line, plane, direction or point being represented. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Sketch first. Name the geometric objects. Use equations only after the relationships are clear. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. At distinction level, algebra and geometry cross-check each other: a parameter value should make sense on the picture, and a geometric condition should guide the equations. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
8. Complex numbers need both algebra and geometry
Distinction standard. A student may manipulate i accurately but still mishandle modulus, argument or locus geometry. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Move repeatedly between rectangular form, polar form and the Argand diagram. Sketch before choosing an argument or interpreting a locus. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. Distinction control appears when conjugates, modulus and argument are chosen for a reason and quadrant information is never outsourced blindly to the calculator. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
9. Differentiation begins with reading the function
Distinction standard. Most differentiation mistakes occur before the derivative is finished: wrong rule, missed composition, poor simplification or failure to interpret the result. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Classify the function structure before differentiating. Decide whether simplification should occur first. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. Then interpret derivatives as rates or local behaviour. This keeps calculus connected to meaning and reduces formula-driven errors. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
10. Optimisation needs the whole feasible problem
Distinction standard. Finding f'(x)=0 is not automatically the end of an optimisation question. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Check the physical or mathematical domain, endpoints where relevant, classification and units. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. Distinction answers return the calculus result to the original problem. A stationary point outside the feasible region is not rescued by flawless differentiation. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
11. Integration should start with preparation
Distinction standard. Strong students do not ask only “which integration formula?” They ask whether the integrand should be simplified, split, substituted, decomposed or rewritten first. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Train a pre-integration pause: inspect structure, identify the likely target form, and estimate what an antiderivative should look like. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. This reduces brute-force attempts and makes later checking through differentiation more natural. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
12. Integration checking should be automatic
Distinction standard. Differentiation provides a powerful independent check for many indefinite integrals. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. After obtaining an antiderivative, differentiate the essential part mentally or symbolically where feasible. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. A-grade checking is selective: use the fastest independent route capable of catching the most expensive error, rather than rereading the same algebra. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
13. Differential equations must preserve modelling meaning
Distinction standard. Separable forms and initial conditions can become mechanical if the variables and constants lose their interpretation. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. State what is changing, separate variables visibly, integrate, preserve the constant and apply conditions before premature approximation. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. Then return to the model and inspect whether the solution’s behaviour is plausible. Distinction control includes the model, not only the differential equation. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
14. Permutations and combinations require outcome definitions
Distinction standard. Choosing nPr or nCr from keywords is fragile. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Define what counts as one outcome and ask whether changing order creates a different outcome. Then check whether cases overlap or omit possibilities. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. Use small examples to audit counting logic. A-grade combinatorics is not fast button selection; it is reliable construction of the sample space. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
15. Probability should be built from one coherent sample space
Distinction standard. Numerator and denominator errors often come from switching reference sets halfway through the solution. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. State the experiment, events and condition before calculating. Use trees, tables or set relations where they make dependence visible. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. The student should be able to explain why the denominator changes under conditioning and why independence is different from mutual exclusivity. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
16. Distribution choice must begin with assumptions
Distinction standard. Binomial and normal commands on the GC make numerical work easy, but model selection is still mathematical. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Write the conditions that justify the model before entering parameters. Identify the random variable and its units. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. A distinction script shows that the probability was modelled correctly, not simply computed accurately. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
17. Sampling distributions need the random variable named
Distinction standard. Students often confuse a population variable with a sample mean or other statistic. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Before formulas, write what is random: an individual observation, a sample mean, or something else. Then derive the relevant mean and variance. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. This prevents automatic copying of population parameters and makes Central Limit Theorem reasoning intelligible rather than ritual. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
18. Hypothesis testing begins before the calculator
Distinction standard. Tail choice, hypotheses and significance level should follow the claim, not the observed result. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Write H0 and H1 about the population parameter before calculating a test statistic or p-value. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. The final conclusion must return to the context. An A-grade statistics answer is mathematically and linguistically aligned with the question being tested. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
19. Correlation and regression need interpretation boundaries
Distinction standard. A strong correlation coefficient does not establish causation, and a regression line is not a law that can be extrapolated indefinitely. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Name explanatory and response variables, identify the data range and distinguish interpolation from extrapolation. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. Distinction-level interpretation uses the model confidently without claiming more than the data support. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
20. Paper movement is a mathematical skill
Distinction standard. A student can know the course and still lose an A through poor allocation of attention. An A target should translate into observable mathematical behaviour, not merely a desire for a high percentage.
Check it. Practise leaving a resistant question before it consumes disproportionate time, marking a return point and restarting cleanly. Use fresh questions and remove unnecessary cues. If the student can succeed only when the chapter or method is announced, the capability is not yet fully portable.
Build it. The goal is not to avoid difficult questions but to protect the rest of the paper. A-grade control means one difficult item cannot collapse the following hour. Then retest after delay or under mixed conditions. The purpose is repeatable control, because examination performance depends on what survives when the surface changes and time becomes finite.
11. Twenty further distinction controls: from strong knowledge to stable marks
21. Routine marks should become boringly reliable
Distinction standard. Distinction performance is built partly by making straightforward and medium-demand marks low-variance. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Take ten reachable questions from different topics and measure not only score but error type and checking success. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. If routine leakage persists, pause the search for harder material and repair the mechanism creating it. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
22. Hard questions should be decomposed before they are attacked
Distinction standard. A difficult H2 problem often contains familiar objects arranged without obvious chapter labels. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Ask the student to identify givens, targets, possible relationships and subproblems before calculating. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Train decomposition until unfamiliarity triggers structure-seeking rather than immediate formula search. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
23. Method comparison should be part of practice
Distinction standard. Strong students can often solve a problem more than one way. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. After one valid solution, ask whether an alternative route would be shorter, more transparent or easier to verify. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. This builds strategic flexibility and helps the student select robust methods under exam conditions. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
24. Exactness should be protected until approximation is useful
Distinction standard. Premature decimalisation can destroy structure, create rounding drift and make later algebra awkward. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Mark points where exact form carries useful information, especially in surds, logarithms, trigonometry and calculus. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Use the GC for numerical support while keeping symbolic structure when the mathematics benefits from it. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
25. Rounding should be a final communication decision
Distinction standard. A student can lose accuracy after correct mathematics simply by rounding too early. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Keep full internal precision and record the required final form or significant figures separately. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. A distinction routine treats rounding as part of answer communication, not as an incidental calculator habit. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
26. Checks should be independent of the original route
Distinction standard. Rereading the same algebra often reproduces the same blind spot. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Use substitution, graph comparison, differentiation of an integral, estimation, boundary checks or units where appropriate. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Choose the cheapest independent check capable of detecting the expensive error. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
27. Question reading should isolate command and condition
Distinction standard. H2 questions can hide important constraints in ordinary language. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Underline or mentally tag what is given, what is required, and any domain, interval, parameter or modelling condition. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. A-grade reading is selective rather than decorative: only information that changes the mathematical route needs active marking. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
28. Time allocation should follow mark opportunity
Distinction standard. Spending twenty minutes protecting one elegant solution can sacrifice easier marks elsewhere. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Track actual time against marks on practice papers and note where the student becomes trapped. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Build a leave-and-return threshold that protects the paper while preserving a route back to difficult work. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
29. Recovery after an error should be rehearsed
Distinction standard. Students often lose more marks from the emotional aftermath of one bad question than from the original mistake. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. In timed sets, deliberately move on after a failed attempt and practise restarting on a new problem. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Distinction control includes the ability to reset cognitive state, not just solve when everything begins smoothly. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
30. Paper 1 and Paper 2 need different mental preparation
Distinction standard. Paper 1 is pure Mathematics, while Paper 2 adds a large probability-and-statistics component. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Practise the transitions in Paper 2 rather than treating the whole course as one undifferentiated bank. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. The student should be able to switch from symbolic pure work to statistical modelling without carrying the wrong habits across. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
31. Statistics conclusions should be rehearsed as mathematics
Distinction standard. A correct p-value with a vague conclusion leaves marks exposed. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Practise short context-specific conclusion sentences tied to the hypotheses and significance decision. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Language should express the statistical claim precisely without overstating evidence. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
32. GC routines should be standardised
Distinction standard. Repeated searching through menus wastes time and increases input errors. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Create a small personal workflow for common graphing, distribution and regression tasks, then practise it under mixed conditions. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. The workflow should still include prediction and verification so speed does not become blind automation. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
33. Formula-list familiarity should reduce search, not replace memory
Distinction standard. An official formula sheet is support, but constant searching interrupts method selection. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Know where major formulas live and which essential facts are not supplied. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Use the list as confirmation while building enough retrieval that routine work remains fluent. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
34. Corrections should become future triggers
Distinction standard. An error record is useful only if it changes behaviour next time. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Convert each recurring error into one question the student asks themselves at the critical point. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Examples include “what is the domain?”, “which variable am I predicting?”, or “does order matter?”. Internal prompts are the bridge from feedback to self-regulation. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
35. Mixed review should include old topics every week
Distinction standard. A student can appear excellent while current material is fresh and still lose earlier methods. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Include a small number of older questions in every weekly set, not only before examinations. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Spacing turns the whole course into one active system rather than a sequence of completed chapters. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
36. Unseen questions should be used as diagnosis, not theatre
Distinction standard. An unfamiliar question is valuable because it reveals recognition and transfer. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. After the attempt, identify exactly where entry failed: reading, representation, method selection or execution. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Do not respond to every unseen failure by simply finding an even harder question; repair the decision that broke. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
37. School tutorials should remain the alignment spine
Distinction standard. External resources can help, but excessive parallel materials fragment attention. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Use current school lectures and tutorials to define sequence, then add external explanation or practice for named gaps. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. A-grade preparation is easier when resources have roles instead of competing for ownership of the syllabus. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
38. Prelim papers should be analysed before being collected
Distinction standard. A folder of difficult school papers creates little value if scripts are not mined for error patterns. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. After each paper, classify lost marks by topic and mechanism, then choose a small repair queue. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. The next paper should test whether that queue changed, not merely add another score to the folder. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
39. High scores should still produce targeted repair
Distinction standard. An 85 or 90 can hide a recurring weakness that later difficult papers exploit. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Review the highest-value lost marks and near-misses even when the grade is strong. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Top-end improvement often comes from removing rare but expensive failure modes, not from increasing ordinary volume. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
40. The final distinction is independence
Distinction standard. A student who can perform only with live tutor prompts has not yet built exam-ready control. The important question is whether this behaviour remains available when the question is unfamiliar, the paper is timed and no adult is present to confirm the route.
Check it. Measure how often the tutor names the method, supplies the representation or confirms a step. Use representative work rather than artificial trick questions. A good check isolates the control you are trying to measure while keeping the Mathematics aligned with the course.
Build it. Over time, those prompts should shrink. The strongest A-grade preparation leaves the learner able to diagnose and recover alone. Then revisit it across more than one paper or week. Stable distinction performance is a repeated state, not a one-day peak.
12. A distinction week: what the training rhythm should contain
A strong weekly system is smaller than many students imagine. It needs current learning, independent tutorial attempts, targeted repair, cumulative retrieval, mixed transfer and some timed control. The exact hours vary with school load, but the functions should all appear. If an entire week is spent only consuming lectures and reading solutions, generation is missing. If the week is only full papers, weak concepts may never be repaired.
One useful rhythm is: learn the school material; attempt before viewing full solutions; mark and classify the first wrong decision; repair the highest-spread errors; revisit one or two older topics; finish with a short mixed set. Add longer timed papers according to the school calendar and examination horizon. This creates a loop rather than a pile of tasks.
The rhythm should be sustainable. A distinction plan that destroys sleep, other subjects or the student’s capacity to think is not a strong plan. Workload is part of performance engineering.
13. What changes at six months, twelve weeks and four weeks from a major examination
Six months out
The main task is syllabus integrity. Repair active prerequisites, finish current topics properly, build mixed retrieval and prevent the error log from growing faster than it shrinks. Full papers are useful, but they should not replace topic repair. The student still has time to change how the system works.
Twelve weeks out
Integration becomes central. Increase mixed work, full-paper exposure and paper-specific control. Use prelims or representative school papers to identify high-value leaks. Keep statistics active and maintain old pure topics. The student should know which three or four mechanisms are still costing the most marks.
Four weeks out
Do not rebuild the whole course. Protect stable knowledge, repair recurring high-value errors, rehearse full-paper movement, maintain sleep and use short targeted sets between papers. The last month is about stabilising the system under examination conditions, not proving bravery with impossible workload.
14. How parents should read an H2 Mathematics score
A percentage is a compressed output. Open it. How many marks were lost because the concept was unknown, because the method was not recognised, because algebra failed, because the GC was mishandled, because statistics language was incomplete, because time ran out, or because checking never happened? Two students with the same 60 may need completely different help.
For high-performing students, the same applies. A move from 82 to 86 may represent meaningful reduction in hard-question leakage. A move from 90 to 93 on a very familiar paper may tell less about transfer. Use several scripts and changed tasks before changing the whole plan.
The page Did the Student Actually Learn? explains this evidence layer in depth.
15. What not to do when aiming for an A
Do not chase a rumoured fixed raw-mark boundary. Do not postpone Probability and Statistics until late revision. Do not collect ten resource libraries. Do not treat every wrong answer as a need for harder questions. Do not use the graphing calculator without prediction. Do not copy corrections without retesting. Do not restart the entire syllabus after one bad paper. Do not turn an A target into a daily judgement of the student’s worth.
The grade target should organise training, not contaminate it. The student needs a calm system that makes high-quality decisions more likely under time.
16. Frequently asked questions
What score guarantees an A in H2 Mathematics?
This page does not invent or promise a fixed raw-score threshold. Grade outcomes depend on the examination process and official awarding. Train the controllable system: knowledge, transfer, accuracy, communication and paper control.
How many prelim papers should I do?
Enough to expose and retest real error patterns. Ten papers with shallow marking can be less valuable than five papers with disciplined analysis and targeted repair. Increase volume only when the feedback loop remains intact.
Should I focus more on Pure Mathematics or Statistics?
Follow the actual 9758 paper structure and your own weaknesses. Pure Mathematics spans both papers, while Probability and Statistics carries substantial weight in Paper 2. Neglecting either creates a structural ceiling.
Do I need very hard questions to get an A?
You need enough unfamiliar and high-demand work to train transfer and problem solving, but routine and medium marks must also be highly reliable. Difficulty should be added after the floor is stable, not used to hide leakage.
How important is the graphing calculator?
Important, but subordinate to mathematical judgement. Use it fluently for appropriate tasks while understanding its limitations, preserving exact form when needed and showing mathematical working when required.
What if I understand lessons but cannot do tutorials alone?
That is a recognition-to-generation gap. Attempt before viewing full solutions, identify the first point of failure, use the smallest useful hint, then redo after the hint is removed. Independent starts are a trainable capability.
What if I keep getting around the same grade despite more work?
Open the score by mechanism. More volume will not help if the active bottleneck is method selection, a recurring prerequisite, time control or uncorrected statistics language. Change the repair, not merely the quantity.
17. Official source and next routes
The examination structure and assessment-objective framework used here come from the 2027 H2 Mathematics 9758 syllabus. Students should always use the syllabus year that applies to their own cohort and their school’s current examination schedule.
For topic-specific learning, use the H2 Mathematics 9758 topic map. For common failure patterns, use H2 Mathematics Common Mistakes. For the first JC term, use H2 Mathematics First Three Months. The separate prelims plan covers the time-bounded twelve-week conversion into full-paper control.
18. The shortest useful answer
An A in H2 Mathematics is not built from one magic technique. It comes from making ordinary marks reliable, keeping assumed A-Math knowledge active, understanding H2 concepts with conditions, selecting methods without chapter labels, protecting statistics, using the GC intelligently, writing enough mathematical evidence, correcting recurring errors and controlling two long papers under time.
The training question is therefore not “How do I force an A?” It is “Which part of my H2 Mathematics system is still too fragile to produce distinction-level work repeatedly?” Find that part, repair it, retest it and keep moving. High grades are outcomes. Reliable mathematical control is the thing you can build.
19. Fifteen worked distinction journeys: how the bottleneck changes the plan
1. From 55 to distinction range: algebra was the hidden tax
Starting state. The student understood new H2 ideas in class but lost marks in rearrangement, factorisation and exact manipulation. More H2 worksheets initially increased frustration because every topic inherited the same symbolic instability. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. The repair block isolated algebraic fractions, exact form and equation manipulation for short daily work, then immediately embedded them back into functions, calculus and vectors. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. The important movement was not simply a later score rise. Tutorial completion became faster, fewer prompts were needed and harder questions could finally be attempted with attention available for the H2 concept. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
2. From strong tutorials to weak tests: chapter labels were doing the selection
Starting state. The student produced excellent topical homework but stalled in mixed assessments. Once told “use integration by parts” or “this is a normal distribution question”, execution was strong. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. Training shifted from more topical volume to mixed classification. Before solving, the student had to name the structure, candidate methods and one condition that justified the choice. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. Performance improved when method recognition, not calculation, became the target. The lesson is that an A system must measure selection separately from execution. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
3. From 80s to low 90s: checking became an independent route
Starting state. The student’s conceptual knowledge was already strong. Lost marks came from signs, copied constants, wrong GC entries and a few unchecked statistical conclusions. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. Instead of adding harder content immediately, practice required one independent verification route on selected questions: substitution, differentiation of an integral, graph comparison, estimation or context check. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. The gain came from converting near-misses. At the top end, removing expensive low-frequency errors can matter more than adding another set of advanced techniques. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
4. Strong Pure Mathematics, weak Paper 2: statistics was treated as revision content
Starting state. The student devoted most of the year to calculus, vectors and complex numbers because those topics felt more intellectually demanding. Probability and statistics were repeatedly postponed. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. The repair made statistics a weekly strand: model assumptions, distribution choice, hypothesis statements, GC workflows and contextual conclusions. Short retrieval was mixed into ordinary weeks rather than saved for the final month. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. Paper 2 became more stable because statistics stopped being a separate late-season project. Distinction preparation must reflect the actual assessment structure, not the student’s favourite part of Mathematics. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
5. Fast calculator, fragile judgement: technology was leading the mathematics
Starting state. The student could use numerical solvers and regression menus quickly but accepted outputs without predicting roots, checking graph windows or preserving exact form. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. Training required a prediction before calculator use and one interpretation after. For graph work, the student had to identify likely features first. For statistics, variables and model assumptions were written before entering commands. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. The GC became faster and safer because mathematical judgement framed every use. Distinction-level technology reduces labour without outsourcing reasoning. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
6. Too many resources, too little closure
Starting state. The student had school notes, two tuition sets, several online channels, shared prelim drives and multiple summary books. Every weak topic generated another resource. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. The system was reduced to school materials as the alignment spine, one supplementary explanation source, one extra problem source and an error ledger. New resources were added only for named gaps. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. Study time shifted from searching and switching into attempts, corrections and retesting. An A plan needs enough information, not maximum information. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
7. Hard-question obsession, routine leakage
Starting state. The student believed an A required constant exposure to the hardest prelim questions and spent most revision time there. Meanwhile, ordinary algebra, notation and rounding errors continued to cost reachable marks. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. Practice was split into reliability and stretch. Routine and medium-demand questions were used to drive error leakage down; difficult questions were retained for transfer and ceiling work. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. The score became more stable because the floor rose. Hard questions matter, but distinction performance collapses if ordinary marks remain volatile. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
8. Full papers every day, no repair loop
Starting state. The student completed large numbers of papers and tracked percentages, but the same mistakes reappeared. Marking produced information without changing later behaviour. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. The new rule was one paper followed by a repair queue: classify first wrong decisions, choose at most three active error families, run targeted work, then use the next paper as a retest. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. Paper volume fell while learning density rose. A full paper is useful only when its evidence changes the next week of practice. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
9. Strong knowledge, unfinished papers
Starting state. Untimed solutions were accurate, yet both papers ended with large blocks incomplete. The student responded by trying to write faster everywhere. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. Timing analysis showed several long traps rather than universal slowness. Training focused on recognising diminishing returns, leaving resistant items, marking return points and protecting easier later marks. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. The student did not become a fundamentally faster thinker overnight. Paper movement improved, and available knowledge had more chances to earn marks. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
10. One bad prelim triggered panic
Starting state. A difficult school prelim produced a sharp score fall, leading the family to consider restarting large parts of the syllabus. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. The script was reopened by mechanism. Several topics remained sound; the main losses were two unfamiliar modelling questions, one late-paper timing collapse and a cluster of statistical communication errors. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. The repair stayed narrow. This protected stable knowledge and used the prelim as diagnosis rather than as a verdict on the entire course. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
11. One excellent prelim triggered overconfidence
Starting state. A very strong paper led the student to stop cumulative retrieval and focus only on rare challenge problems. Earlier methods became less accessible over the following weeks. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. The plan restored spaced mixed review while keeping extension work. High scores were treated as evidence of current performance, not permission to retire the course. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. Distinction stability depends on maintaining the system that produced the peak. A high result should simplify priorities, not dismantle maintenance. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
12. A-Math strength hid H2 statistics weakness
Starting state. The student entered JC with excellent algebra and calculus, so early H2 performance was strong. Later, sampling and hypothesis testing exposed a different kind of weakness: assumptions and contextual interpretation. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. Training shifted from symbolic fluency to explicit modelling language. The student stated random variables, parameters, hypotheses and conclusions in context before using the GC. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. The case shows why prior A-Math success is a runway, not a complete H2 profile. Distinction requires both pure and statistical modes of thinking. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
13. Good explanations, weak independence
Starting state. The student could follow every tuition explanation and reproduce methods during the lesson but could not start comparable homework later. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. Prompts were delayed and recorded. The tutor moved from full explanations to smaller cues, then to fresh delayed questions without method labels. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. The first lesson performance looked worse, but independent starts rose. An A system should reward what the learner can carry alone, not how polished the guided lesson appears. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
14. Accurate but slow: fluency was the missing layer
Starting state. The student selected methods correctly and rarely made conceptual errors, yet algebraic execution and GC workflows consumed too much time. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. Short timed sets were introduced only on already-secure skills. The objective was not rushing; it was reducing unnecessary cognitive effort in routine steps while protecting accuracy. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. Speed improved because fluent components became cheaper. Timed practice works best after the mathematical route is stable enough to automate safely. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
15. Confident but inconsistent: evidence replaced mood
Starting state. The student alternated between “I’m definitely getting an A” and “I’m terrible at H2 Maths” depending on the latest test. The first task is to describe the system accurately before increasing workload. A grade target does not tell us which capability is limiting the student.
Change. A small dashboard tracked independent starts, recurring errors, mixed-set accuracy, paper completion and delayed retrieval across several weeks. The intervention is deliberately matched to the bottleneck. This keeps revision efficient and makes the next piece of evidence interpretable.
What changed. Confidence became tied to observable control rather than one score. Distinction preparation benefits when emotion is informed by evidence instead of driving the plan. The principle is transferable: improve the mechanism that is currently leaking marks, then retest it under conditions closer to the examination.
20. A final distinction audit before major examinations
Before the final examination phase, audit five things. First, can the student retrieve assumed and H2 knowledge after delays? Second, can they select methods in mixed work without chapter labels? Third, are recurring errors shrinking rather than simply being recopied into an error book? Fourth, can both papers be completed with deliberate movement and enough time for targeted checking? Fifth, can the student recover after one difficult question without carrying the disruption into the rest of the script?
If any answer is no, choose the smallest repair that changes it. The final weeks are not a referendum on intelligence. They are a period of controlled stabilisation. Protect sleep, current strengths and the routines that already work while reducing the few high-value failure modes that remain.
A distinction system is successful when it becomes quieter near the examination, not more frantic. The student knows what to revise, which errors matter, how to use the GC, when to move on, what to check and how to return to a difficult problem. The paper can still be hard. Control is the difference.