Exam guide

1. Paper structure

Changes over 2022–2026

• 2022 (May/June and Oct/Nov) were the last sessions of the old syllabus. Paper 2 was out of 50, and its Section A (about 20 marks) asked about pre-release material, a scenario students programmed before the exam. Paper 1 also covered HTML structure and presentation, plus ethics and software licensing (freeware, free software, shareware).
• From 2023 the new syllabus applies. There is no pre-release material. Paper 2 is out of 75 and always ends with a 15-mark extended program, written in pseudocode, Python, Java or VB.NET. New Paper 1 content: automated systems, robotics, AI and expert systems, digital currency, packet switching, encryption, two's complement and binary shifts. Logic gates moved to Paper 2.
• From 2026 the syllabus allows functions with up to three parameters (it was two). The change came after the June 2025 Paper 22 final question used three-parameter functions and was voided (see section 7).
• The number of parts keeps growing: Paper 12 had 21 parts in June 2022 and 38 in June 2026. Expect many short 1–2 mark items.

2. Topic weights

Shares are of all 1,300 sampled marks. The new-syllabus column covers 2023-MJ to 2026-MJ only (1,050 marks). "Sessions examined" counts sessions out of 9.

Within each paper (new syllabus):

• Paper 1: Hardware 25%, Data representation 24%, Internet and its uses 19%, Data transmission 13%, Automated and emerging technologies 10%, Software 10%.
• Paper 2: Algorithm design 42%, Programming 36%, Boolean logic 11%, Databases 11%.

Biggest subtopics by marks:

• the 15-mark extended program: 105 marks;
• number systems and conversions: 66;
• internet, WWW, DNS and cookies: 60;
• security solutions: 55;
• logic circuits and expressions: 55;
• storage: 55;
• validation and verification: 53;
• trace tables: 50.

3. Bankers

Examined in every session, or every new-syllabus session:

• Number systems: binary, denary and hexadecimal conversions, and where hex is used (9/9).
• Binary addition, overflow, logical shifts and two's complement (7/7 new-syllabus sessions).
• Storage: primary vs secondary; magnetic, optical and solid state; virtual memory and cloud storage (9/9).
• Security solutions: firewalls, proxy servers, SSL/TLS, access levels, backups (9/9). Cyber threats come alongside them, for example phishing, DDoS, malware and brute force.
• Internet and WWW: URLs, HTTP/HTTPS, how DNS finds a web page, browser functions, cookies (8/9).
• CPU: components (ALU, CU, registers), the fetch–decode–execute cycle and performance (7/9).
• AI and expert systems: knowledge base, rule base, inference engine, machine learning (7/7 new).
• Validation and verification: range, length, type, presence, format and check-digit checks; double entry and visual check (9/9).
• Trace tables (9/9) and finding and correcting errors in pseudocode (9/9).
• Standard methods: totalling, counting, max/min/average, linear search, bubble sort (7/7 new).
• Databases: primary key, data types, and completing or reading an SQL SELECT … FROM … WHERE … ORDER BY query (9/9).
• Logic: draw a circuit or write an expression from a statement, then complete the truth table (9/9).
• The 15-mark program at the end of Paper 2 (7/7 new): arrays, loops, validation, totals, averages and clear output.

4. Rarely tested

Each of these appeared in only one or two sessions, so a question on them is overdue rather than impossible:

• Interrupts: only 2025-ON.
• Digital currency and blockchain: only 2023-ON.
• Automated systems (sensors, microprocessor and actuators in industry, transport and so on) as a named question: only 2023-ON. Sensors and control did appear in 5 sessions.
• Packet switching and packet structure: 2024-MJ and 2026-MJ only.
• Symmetric and asymmetric encryption: 2024-MJ and 2026-MJ only.
• Network hardware and addressing (NIC, MAC, IPv4/IPv6, router): 2023-MJ and 2024-MJ only.
• Robotics: only 2025 (both sessions).
• Maintainable programs (comments, meaningful names, procedures) as a separate question: only 2023-MJ. These are still marked inside every 15-mark program.
• File handling: 3 sessions. Procedures, functions and parameters: 3 sessions. Both are likely to rise now that the 2026 syllabus allows three parameters.

5. Trends

• Paper 2 now leans on algorithm design. Its marks went from about 20 per session in 2022 to 25–40 since 2023 (34 in June 2026), through more trace tables, error-finding, flowchart completion and "write an algorithm" items.
• "The internet and its uses" is shrinking. It fell from 30 marks in 2022-MJ (old syllabus with HTML) to 10–14 marks per session since 2024.
• Data transmission is rising. It was 4–5 marks in 2022–2023 and 15 marks in June 2026: packet switching, parity bytes, error detection and encryption together.
• Emerging technologies are now regular, usually 4–11 marks: AI and expert systems almost every time, robotics in 2025.
• Context questions are growing. Paper 1 hangs most questions on a scenario (a smartwatch, a video streaming service, a flower-identifying app, a surgery robot), and examiners reward answers tied to it.
• The final program has run on scenarios such as banking log-ins (2023-MJ), flooring quotations (2023-ON), a grid game (2024-MJ), room areas (2024-ON), volume functions (2025-MJ), counting and sorting numbers (2025-ON) and daily rainfall (2026-MJ). It nearly always needs 1D or 2D arrays, loops, input validation and formatted output.

6. Question styles

• Paper 1:
  • The most common command words are Give, Explain, Identify, State and Describe.
  • Typical formats: tick-box tables (true/false statements, input or output, which storage), "complete the table of terms and descriptions", fill-the-gap paragraphs, and number conversions or binary additions with "show your working".
  • Explain questions worth 4–6 marks want a step-by-step process, for example how DNS finds a page, how a firewall filters traffic, or how an expert system reaches a diagnosis.
• Paper 2:
  • The most common command words are Write, Complete, State and Identify.
  • Typical formats:
    • completing a trace table from a flowchart or pseudocode;
    • identifying 3–5 errors by line number and correcting them;
    • test data (normal, abnormal, extreme, boundary);
    • drawing logic circuits and completing truth tables;
    • completing an SQL query;
    • writing short pseudocode, such as a double-entry check or an average with ROUND;
    • the 15-mark program.
• Answer in the form the question asks for: pseudocode when it says pseudocode, a flowchart when it says flowchart. Use the syllabus pseudocode style, for example ←, ENDIF, NEXT Counter, SUBSTRING, LENGTH, UCASE, ROUND.

7. Common mistakes

These are paraphrased from the examiner reports for June 2022, June 2023 and June 2025 (Papers 12 and 22).

• Generic answers that ignore the context. Examples: lossless compression of a text file explained as if it were an image, or the reverse; robot advantages that ignore the human operator in the question; a firewall described in general instead of for the situation given.
• Repeating what the question already says, for example giving RAM as "another" primary memory when RAM was the example given, or "easier to read" when the question already said it.
• Vague number-system reasons. "Hex saves storage space" is wrong. Hex is shorter to read and write for humans; the computer still stores binary. When a hex digit is 0, still write the 0.
• Two's complement: giving the binary for +22 when −22 was asked.
• Binary working: leaving out the working when the question says "show all your working" costs marks.
• Confusing high-level and low-level languages, and describing a compiler's benefits when asked how it works (it translates the whole program and produces an executable or an error report).
• IDE functions: restating the name ("auto-complete completes code") instead of describing what it does. Mixing up OS functions with IDE functions.
• CPU: writing "clock speed" when the component is the "clock".
• Embedded systems: "does one task" alone is not enough.
• Expert systems: naming the parts but not explaining how the inference engine applies the rules to the facts to reach a conclusion.
• SSL/DNS processes: too little technical detail. Mention digital certificates, encryption, the IP address returned by DNS, and the web server sending the page.
• Verification vs validation: verification checks that data was not changed during input or copying. It does not check that data is correct. Do not add range checks when only a double-entry check is asked for.
• SQL: forgetting quote marks around text values such as 'Asia'.
• Pseudocode: writing code that only half-follows the given variable names, or writing a non-pseudocode style. In the 15-mark program, students lose time declaring things not needed and miss requirements stated at the end of the scenario. Comments and input and output messages are marked.
• Logic: putting a NOT gate in the wrong place in a circuit, for example before an XOR instead of after it.
• Voided questions: June 2023 Paper 22 Question 5 was removed from marking, and June 2025 Paper 22 Question 10 (three-parameter functions) was given full marks to all candidates.

8. Grade thresholds

The table gives syllabus thresholds for option AY (Papers 12 + 22). The maximum is 125 in 2022 (75 + 50) and 150 from 2023 (75 + 75).

The next table gives component thresholds (A / C) for Paper 12 and Paper 22:

What the thresholds show:

• On the new 150-mark papers, a C has needed about 40–54 marks (27–36%), an A 84–100 (56–67%), and an A* 104–123 (69–82%).
• June 2025 was unusually high. Part of the reason is that Paper 22 Question 10 (15 marks) was credited in full to everyone.
• Thresholds differ by variant, and options AX and AZ have their own rows in the source files.

9. Revision priorities for Oct/Nov 2026 and 2027

1. Practise a 15-mark program every week, timed at about 30 minutes. Include arrays (1D and 2D), a menu or loop, validation, totals and averages, functions with up to three parameters (new from 2026), comments and clear messages. It is the single largest block of marks.
2. Trace tables and error-finding: do one of each per paper, and work line by line in ink. Together they are worth about 85 marks over the sample.
3. Validation, verification and test data. Know every check by name and be able to code a range check and a double-entry check in pseudocode.
4. Number work on Paper 1: binary, denary and hex conversions, binary addition with overflow, logical shifts, two's complement, and file-size calculations (bits, bytes, KiB, MiB).
5. Process explanations:
   • the fetch–decode–execute cycle;
   • how DNS returns a web page;
   • how SSL/TLS works;
   • how parity, checksum, echo check and ARQ work;
   • how an expert system reaches a decision;
   • how lossless (RLE) and lossy compression work for text, images and sound.
6. Security: pair each threat (phishing, pharming, DDoS, brute force, malware types, social engineering) with its solution. Practise applying them to a given scenario.
7. Logic and SQL: drawing circuits from statements, truth tables, and SELECT/WHERE/ORDER BY with AND/OR, SUM and COUNT.
8. Catch up on the overdue topics: interrupts, digital currency and blockchain, packet structure, symmetric and asymmetric encryption, IPv4 vs IPv6 and MAC addresses, robotics, and file handling.

Comparison with O Level Computer Science (2210)

• Same question papers. We compared the text of the variant 2 papers for 0478 and 2210 from Oct/Nov 2022 to June 2026. Paper 12 and Paper 22 had the same questions in every session; only the cover page, barcodes and page layout differ. A student revising for either qualification can use both sets of past papers freely.
• Differences:
  • 0478 sits three variants per session (11–13 and 21–23). The 2210 files have two (11–12 and 21–22).
  • The grade thresholds are set separately. Check the right gt file for your syllabus code.
  • The local 2210 sample starts at Oct/Nov 2022. This 0478 sample also includes June 2022, the old-syllabus pre-release paper.