Cambridge practical and extended reasoning

AI-assisted English content review; independent teacher approval pending.

Practice scope and review

Original, AI-authored English practice with worked guidance. Selected topics only; independent teacher review is pending. This is not a complete syllabus, official past-paper collection or predicted exam score. Check your own exam-board specification: these concepts are not certified against a particular board or exam year.

160 original questions across 22 selected topic areas. No account or payment required to practise.

Cambridge International AS & A Level Computer Science (9618), examination year 2027: AI-assisted conceptual alignment to official sources, not teacher or exam-board certification. Syllabus mapping and remaining coverage.

5 questions in this topic

Exact English revisions checked in an AI-assisted academic review; independent teacher approval is still pending.

Choose practice topic

All topics

Original multi-stage supplied-data practice. Partial written support only, not complete syllabus coverage, official results or practical certification.

5 of 5 questions shown

Loading saved responses

Question 1Hide question5 indicative marks

Calculate, analyse and evaluate5 indicative marksNo calculator

Supplied fictional sensor inputs are [0, 10, 100]. A draft mean_valid uses '0 < value < 100' and divides total by count with no empty check. Contract: accept integer readings from 0 to 100 inclusive (reject booleans), ignore all other entries, and return None if none qualify. Trace the corrected total/count after every input and give the return. Supply runnable Python 3 code and tests for boundaries, normal, abnormal, empty and all-invalid inputs; explain which tests expose each draft fault and what an honest execution-evidence record would contain. This is conceptual practice, not candidate Paper 4 execution. Give a connected worked response with calculations, justified conclusions and limitations. All datasets are original fictional practice, not official experiment results. Written/supplied-data work only; any actual laboratory or field activity requires teacher approval, risk assessment and supervision. No human trials. Primary reading: https://www.cambridgeinternational.org/Images/721397-2027-2029-syllabus.pdf

Cambridge 2027: Partial conceptual match

Runnable console model, trace and test evidence: selected written evidence only. Published Python block executed by regression test with additional independent oracle cases and a boundary-mutant check; learner execution explicitly unproven. Science practical-N and computing rN rows are internal identifiers, not official numbered leaf syllabus codes. Aggregate rows are only partially supported. Actual apparatus use, candidate coding execution and practical attainment are not demonstrated here. Independent teacher review remains required.

9618: 11.3 Structured Programming / Candidates should be able to / objective row 4: Define and use functions
9618: 12.3 Program Testing and Maintenance / Candidates should be able to / objective row 3: Correct detected errors
9618: 12.3 Program Testing and Maintenance / Candidates should be able to / objective row 5: Test strategy and test plan contents
9618: 12.3 Program Testing and Maintenance / Candidates should be able to / objective row 6: Choose normal, abnormal and boundary test data

AI-assisted placement. Independent teacher approval pending.

Your response to question 1

Show indicative guidance

Worked guidance

For 0 the total remains zero but count increases to one; 10 gives total 10/count 2; 100 gives total 110/count 3. The draft would accept only 10 and return 10, so a normal middle-value test alone misses its boundary fault. Empty and all-invalid cases expose its unguarded division.

import unittest

def mean_valid(values):
    total = 0
    count = 0
    for value in values:
        if type(value) is int and 0 <= value <= 100:
            total += value
            count += 1
    return total / count if count else None

class MeanTests(unittest.TestCase):
    def test_normal(self):
        self.assertEqual(mean_valid([20, 40]), 30)
    def test_boundaries(self):
        self.assertEqual(mean_valid([0, 100]), 50)
        self.assertAlmostEqual(mean_valid([0, 10, 100]), 110 / 3)
    def test_abnormal(self):
        self.assertIsNone(mean_valid([-1, 101, True, 3.5, '5', None]))
    def test_empty(self):
        self.assertIsNone(mean_valid([]))

if __name__ == '__main__':
    unittest.main()

A plausible alternative is a filtered list plus sum/len with the same explicit type and empty rules, trading extra storage for brevity. Capture the exact code revision, environment, test inputs, expected/actual returns and pass/fail output, then document any correction and regression rerun. Do not fabricate a screenshot or say a learner ran this because the author tested it. A small passing suite does not prove correctness for arbitrary iterables or establish practical assessment competence.

Self-marking points

Tick what your working earned. The count is your own self-assessment.

Corrected states are (0,1), (10,2), (110,3); result is 110/3, about 36.6667.
Inclusive comparisons fix boundary exclusion; an empty guard avoids division by zero.
Reject bool explicitly because Python bool is an int subclass; specify behaviour for strings, floats and None.
Use independent expected results and retain actual outputs, code version and test identity, including failures.
Author-side execution is not learner execution, debugging or submitted Paper 4 evidence; practical competence remains on hold.

Self-assessed: not counted yet. This question is worth 5 marks.

Examiner tip

Indicative original worked practice, independent teacher approval pending. Explain assumptions and alternatives; do not equate a correct written answer with observed practical performance.

Question 2Fictional import contract: an ID must be exactly three ASCII digits (leading zeros allowed), a reading must be an integer 0..100, and duplicate IDs are rejected within the batch. Records are ('007',0), ('008',100), ('007',50), ('009',101), ('01A',25). Work through acceptance in order, identify checks and design boundary tests. The source document says 50 for ID 008. Explain why a valid imported 100 is still wrong, and compare visual/double-entry verification with a checksum. No real personal data are used. Give a connected worked response with calculations, justified conclusions and limitations. All datasets are original fictional practice, not official experiment results. Written/supplied-data work only; any actual laboratory or field activity requires teacher approval, risk assessment and supervision. No human trials. Primary reading: https://www.cambridgeinternational.org/Images/721397-2027-2029-syllabus.pdf5 indicative marks

Calculate, analyse and evaluate5 indicative marksNo calculator

Fictional import contract: an ID must be exactly three ASCII digits (leading zeros allowed), a reading must be an integer 0..100, and duplicate IDs are rejected within the batch. Records are ('007',0), ('008',100), ('007',50), ('009',101), ('01A',25). Work through acceptance in order, identify checks and design boundary tests. The source document says 50 for ID 008. Explain why a valid imported 100 is still wrong, and compare visual/double-entry verification with a checksum. No real personal data are used. Give a connected worked response with calculations, justified conclusions and limitations. All datasets are original fictional practice, not official experiment results. Written/supplied-data work only; any actual laboratory or field activity requires teacher approval, risk assessment and supervision. No human trials. Primary reading: https://www.cambridgeinternational.org/Images/721397-2027-2029-syllabus.pdf

Cambridge 2027: Partial conceptual match

Validation, verification and a false reassurance: selected written evidence only. Ordered acceptance, preserved leading zeros, accepted aggregate and boundary outcomes independently checked; validity separated from truth. Science practical-N and computing rN rows are internal identifiers, not official numbered leaf syllabus codes. Aggregate rows are only partially supported. Actual apparatus use, candidate coding execution and practical attainment are not demonstrated here. Independent teacher review remains required.

9618: 6.2 Data Integrity / Candidates should be able to / objective row 1: Validation and verification protect integrity differently
9618: 6.2 Data Integrity / Candidates should be able to / objective row 2: Range, format, length, presence, existence, limit and check-digit validation
9618: 6.2 Data Integrity / Candidates should be able to / objective row 3: Visual/double-entry verification; parity and checksums on transfer
9618: 12.3 Program Testing and Maintenance / Candidates should be able to / objective row 6: Choose normal, abnormal and boundary test data

AI-assisted placement. Independent teacher approval pending.

Your response to question 2

Show indicative guidance

Worked guidance

Treat IDs as strings so '007' retains its zeros. Require length three and each character between '0' and '9', rather than a broad Unicode digit predicate that accepts characters outside this contract. Maintain a set of accepted IDs; this batch gives two acceptances followed by three documented rejections. Mean accepted reading is (0 + 100)/2 = 50.
Test readings -1, 0, 1, 99, 100, 101, an empty value, a string and a Boolean; test IDs '07', '007', '0007', '01A' and a repeated valid ID. State expected rejection reasons before running tests so a crash is not mistaken for controlled validation.
The imported 100 passes range validation but disagrees with source 50. A source comparison or independent re-entry can detect transcription error; two identical entries do not guarantee the source itself is accurate. A transfer checksum compares encoded data integrity and admits collisions, so it is no substitute for source verification. Keep rejected records and correction provenance instead of silently changing them. Aggregate plausibility and passing test cases remain limited written evidence.

Self-marking points

Tick what your working earned. The count is your own self-assessment.

Accept only ('007',0) and ('008',100); reject duplicate 007, out-of-range 101 and non-digit 01A.
Accepted count is 2 and mean reading is 50, but this aggregate masks the wrong source value.
Range/format/presence and in-batch uniqueness validation cannot establish agreement with the source.
Verify 008 against its source or independent entry; matching repeated entries can still share a source error.
A checksum can detect some transfer corruption, not semantic correctness or every possible error.

Self-assessed: not counted yet. This question is worth 5 marks.

Examiner tip

Indicative original worked practice, independent teacher approval pending. Explain assumptions and alternatives; do not equate a correct written answer with observed practical performance.

Question 3Simulate only a fictional temperature controller, initially OFF. For each valid reading, turn ON at temperature <=18, turn OFF at >=22, otherwise retain state. Missing input forces OFF and raises a fault; the next valid input clears the fault and applies the rule to that OFF state. Readings / degrees C are 17, 19, 21, 22, missing, 20, 18. Trace state/fault after each, describe state transitions, compare with one threshold at 20, and propose boundary, persistence and missing-data tests. Explain monitoring versus control and why this is not deployable heater safety logic. Give a connected worked response with calculations, justified conclusions and limitations. All datasets are original fictional practice, not official experiment results. Written/supplied-data work only; any actual laboratory or field activity requires teacher approval, risk assessment and supervision. No human trials. Primary reading: https://www.cambridgeinternational.org/Images/721397-2027-2029-syllabus.pdf5 indicative marks

Calculate, analyse and evaluate5 indicative marksNo calculator

Simulate only a fictional temperature controller, initially OFF. For each valid reading, turn ON at temperature <=18, turn OFF at >=22, otherwise retain state. Missing input forces OFF and raises a fault; the next valid input clears the fault and applies the rule to that OFF state. Readings / degrees C are 17, 19, 21, 22, missing, 20, 18. Trace state/fault after each, describe state transitions, compare with one threshold at 20, and propose boundary, persistence and missing-data tests. Explain monitoring versus control and why this is not deployable heater safety logic. Give a connected worked response with calculations, justified conclusions and limitations. All datasets are original fictional practice, not official experiment results. Written/supplied-data work only; any actual laboratory or field activity requires teacher approval, risk assessment and supervision. No human trials. Primary reading: https://www.cambridgeinternational.org/Images/721397-2027-2029-syllabus.pdf

Cambridge 2027: Partial conceptual match

Hysteresis trace and control limitations: selected written evidence only. State/fault trace recomputed including equality and missing recovery; history dependence and non-deployment safety boundary explicit. Science practical-N and computing rN rows are internal identifiers, not official numbered leaf syllabus codes. Aggregate rows are only partially supported. Actual apparatus use, candidate coding execution and practical attainment are not demonstrated here. Independent teacher review remains required.

9618: 3.1 Computers and their components / Candidates should be able to / objective row 8: Monitoring/control, sensors, actuators and feedback
9618: 12.2 Program Design / Candidates should be able to / objective row 2: State-transition diagram purposes
9618: 12.3 Program Testing and Maintenance / Candidates should be able to / objective row 5: Test strategy and test plan contents

AI-assisted placement. Independent teacher approval pending.

Your response to question 3

Show indicative guidance

Worked guidance

At 17 turn ON. Readings 19 and 21 lie inside the band so ON persists. At 22 turn OFF. Missing input sets OFF/fault true; 20 clears the fault but remains OFF inside the band. Finally 18 turns ON. The same temperature 20 may give ON or OFF depending on history, so this is not a stateless Boolean comparison.
Describe OFF -> ON at <=18 and ON -> OFF at >=22 with self-loops in the band; missing data takes either state to OFF/fault. Test exactly 18 and 22, just inside/outside each threshold, long repeated in-band sequences, missing while ON, and recovery at 20. Record expected states before simulation and compare each step, not merely the final state.
With one threshold at 20, noisy readings either side can switch repeatedly; separate thresholds reduce that behaviour. Hysteresis alone cannot detect a sensor stuck at 17 or an actuator stuck ON. Independent limits, fault handling and a context-specific risk assessment would be required by qualified engineers before any real control system. No hardware wiring, mains switching or human temperature trial is requested or authorised.

Self-marking points

Tick what your working earned. The count is your own self-assessment.

State sequence is ON, ON, ON, OFF, OFF, OFF, ON; fault is true only at the missing reading.
Between 18 and 22 the previous valid control state persists; equality boundaries are explicit.
A single threshold can chatter near 20; hysteresis reduces switching but introduces a band.
Monitoring records readings; closed-loop control also uses feedback to command an actuator.
Sensor lag, stale plausible readings and actuator failure require engineering safeguards; this is a simulation, not a safety-approved controller.

Self-assessed: not counted yet. This question is worth 5 marks.

Examiner tip

Indicative original worked practice, independent teacher approval pending. Explain assumptions and alternatives; do not equate a correct written answer with observed practical performance.

Question 4For fictional sorted array [3,7,11,15,19,23,27], binary search uses zero-based inclusive low/high and mid=floor((low+high)/2), comparing the middle value with the target once per iteration. Trace targets 19 and 20, including bounds, and compare linear-search element visits. A supplied benchmark over 10000 identical lookups reports binary 0.030 s and linear 0.020 s. Evaluate the claim 'binary search is slower in general'. Discuss sortedness, absent/edge tests, preprocessing cost and a fair measurement plan; do not claim a benchmark was actually run. Give a connected worked response with calculations, justified conclusions and limitations. All datasets are original fictional practice, not official experiment results. Written/supplied-data work only; any actual laboratory or field activity requires teacher approval, risk assessment and supervision. No human trials. Primary reading: https://www.cambridgeinternational.org/Images/721397-2027-2029-syllabus.pdf5 indicative marks

Calculate, analyse and evaluate5 indicative marksNo calculator

For fictional sorted array [3,7,11,15,19,23,27], binary search uses zero-based inclusive low/high and mid=floor((low+high)/2), comparing the middle value with the target once per iteration. Trace targets 19 and 20, including bounds, and compare linear-search element visits. A supplied benchmark over 10000 identical lookups reports binary 0.030 s and linear 0.020 s. Evaluate the claim 'binary search is slower in general'. Discuss sortedness, absent/edge tests, preprocessing cost and a fair measurement plan; do not claim a benchmark was actually run. Give a connected worked response with calculations, justified conclusions and limitations. All datasets are original fictional practice, not official experiment results. Written/supplied-data work only; any actual laboratory or field activity requires teacher approval, risk assessment and supervision. No human trials. Primary reading: https://www.cambridgeinternational.org/Images/721397-2027-2029-syllabus.pdf

Cambridge 2027: Partial conceptual match

Search traces and benchmark validity: selected written evidence only. Inclusive-bound traces, visit counts and batch-to-lookup unit conversion checked independently; hypothetical benchmark not passed off as execution. Science practical-N and computing rN rows are internal identifiers, not official numbered leaf syllabus codes. Aggregate rows are only partially supported. Actual apparatus use, candidate coding execution and practical attainment are not demonstrated here. Independent teacher review remains required.

9618: 12.3 Program Testing and Maintenance / Candidates should be able to / objective row 4: Select dry-run, walkthrough, white/black-box, integration, alpha, beta, acceptance or stub tests
9618: 19.1 Algorithms / Candidates should be able to / objective row 1: Implement linear/binary search; binary-search preconditions and scaling
9618: 19.1 Algorithms / Candidates should be able to / objective row 5: Compare algorithm time/space using Big O

AI-assisted placement. Independent teacher approval pending.

Your response to question 4

Show indicative guidance

Worked guidance

For 19: bounds (0,6) give mid 3/value 15, so low becomes 4; (4,6) give mid 5/value 23, so high becomes 4; (4,4) give mid 4/value 19, found. For 20 the first two updates match, but value 19 is smaller so low becomes 5 and the loop stops. This counts middle-element visits, not every machine-level comparison.
A linear scan checks indices 0..4 for 19 and all seven for an absent value if it uses equality only. Because the array is sorted, a variant stopping at the first larger value stops at 23 after six visits for 20; state which algorithm is compared. Check empty, single-element, first/last and absent targets, and require sorted input for binary search; duplicates need a declared any-match or first-match contract.
Divide each supplied batch time by 10000: 3.0e-6 s and 2.0e-6 s per lookup. Constants and interpreter overhead can dominate for seven elements. Test many sizes and target distributions with the same environment, exclude or explicitly include preprocessing, and report repeats and spread. Sorting a one-off dataset can outweigh a lookup saving; many lookups may amortise it. The invented timings support only this stipulated comparison, not actual execution evidence or a universal performance conclusion.

Self-marking points

Tick what your working earned. The count is your own self-assessment.

Target 19 uses mids 3,5,4 and is found at index 4 in three visits.
Target 20 uses mids 3,5,4 then terminates with low 5 > high 4, absent in three visits.
Simple equality-only linear search visits five elements for 19 and seven for absent 20; an early-stop sorted variant may visit six for 20.
Supplied per-lookup times are 3 and 2 microseconds; the small example does not override O(log n) versus O(n) scaling.
Use comparable implementations, varied sizes/targets and repeated timed batches; account for sorting, warmup and measurement noise.

Self-assessed: not counted yet. This question is worth 5 marks.

Examiner tip

Indicative original worked practice, independent teacher approval pending. Explain assumptions and alternatives; do not equate a correct written answer with observed practical performance.

Question 5An encoder maps a list of strings to consecutive (value,count) runs; equal values separated by another value must remain separate runs. Trace input ['A','A','B','A']. Give Python 3 code that returns [] for empty input without changing the input. State tests that distinguish run-length encoding from frequency counting, and analyse time and auxiliary space, excluding the returned output. Assume string equality comparisons take constant time.5 indicative marks

Analyse / Explain / Evaluate5 indicative marksNo calculator

An encoder maps a list of strings to consecutive (value,count) runs; equal values separated by another value must remain separate runs. Trace input ['A','A','B','A']. Give Python 3 code that returns [] for empty input without changing the input. State tests that distinguish run-length encoding from frequency counting, and analyse time and auxiliary space, excluding the returned output. Assume string equality comparisons take constant time.

Cambridge 2027: Partial conceptual match

cs9618-2027-v2: RLE compression (PDF15), test-data selection (PDF31), and Big O time/space (PDF37). One Python encoder exercises run representation, empty/boundary traces and complexity under constant-time string comparisons. Partial coverage only: not other compression algorithms, the full testing lifecycle, or assessed Paper4 practical competence.

9618: 1.3 Compression / Candidates should be able to / objective row 3: Compress text, images and sound, including RLE
9618: 12.3 Program Testing and Maintenance / Candidates should be able to / objective row 6: Choose normal, abnormal and boundary test data
9618: 19.1 Algorithms / Candidates should be able to / objective row 5: Compare algorithm time/space using Big O

AI-assisted placement. Independent teacher approval pending.

Your response to question 5

Show indicative guidance

Worked guidance

The first two A values form one run, B ends it, and the last A starts a new run. Counting all A values together would lose ordering and cannot reconstruct the original list.

def encode(values):
    if not values:
        return []
    result = []
    current, count = values[0], 0
    for value in values:
        if value == current:
            count += 1
        else:
            result.append((current, count))
            current, count = value, 1
    result.append((current, count))
    return result

The final append is necessary because no next value arrives to close the last run. Check [] -> [], ['A'] -> [('A',1)], ['A','A'] -> [('A',2)] and the supplied mixed input; compare the input with a saved copy afterwards. Each element is visited once. Only current and count are kept beyond result, giving constant auxiliary space; alternating strings can produce n output pairs. These are test expectations, not a claim of candidate execution evidence.

Self-marking points

Tick what your working earned. The count is your own self-assessment.

Output [('A',2),('B',1),('A',1)].
Initialise only after handling empty input.
Flush each completed run and the final run.
Tests cover empty, singleton, repeated and separated equal values; input unchanged.
O(n) time and O(1) auxiliary space, with O(n) worst-case output space.

Self-assessed: not counted yet. This question is worth 5 marks.

Examiner tip

Support each conclusion with the stated evidence and reasoning. These are indicative self-marking points; other well-supported answers are acceptable. Independent teacher review pending.

All AS & A Level subjects