GCSE计算机 编程评估 机考 AQA OCR Edexcel对比
同样一张 GCSE 计算机科学证书，AQA、Pearson Edexcel 和 Cambridge OCR 对编程技能的评估方式并不一样：两家把代码写在纸上，一家让考生坐在电脑前真写程序。评估媒介的差别直接决定复习时该练什么，也决定考场上时间该怎么分配。这篇文章把三个考试局的规格说明摊开对照，顺便说清 Ofqual 在 2025 年 12 月启动的机考监管咨询会怎样影响后面的考试安排。
The three main boards for GCSE computer science in England do not assess programming in the same way. Two of them ask candidates to write code by hand; one puts them in front of a computer. The rules come from the boards' own specifications and from Ofqual, the regulator, so the practical question for a student is simple: what exactly gets marked, and in what medium.
一、编程技能落在哪张卷子上 / Which paper carries the programming
三个考试局的 GCSE 计算机科学都是两卷制，每卷各占百分之五十，没有计入成绩的课程作业。编程技能不会平均分在两卷里，它集中在一张卷子上。AQA 8525 把它放在 Paper 1，OCR J277 放在 Component 02，Pearson Edexcel 1CP2 放在 Paper 2。这个位置很关键，因为另一张卷子的复习策略完全不同。很多学生以为两卷都要写完整程序，实际上理论卷考的是数据表示、计算机系统与网络这些内容，只有一张卷子会让你设计、写、测试并改进代码。弄清楚哪张卷子负责编程，复习计划才不会两头空。
All three boards split the qualification into two equally weighted papers, and none of them counts coursework towards the final grade. Programming is concentrated in a single paper, and that paper differs by board: AQA 8525 puts it in Paper 1, OCR J277 in Component 02, Pearson Edexcel 1CP2 in Paper 2. Many students revise both papers as though they were the same task; in practice the theory paper rewards recall and explanation, while the programming paper rewards reading code accurately and writing it under time pressure.
二、AQA 8525 的 Paper 1 手写代码 / Writing code by hand in AQA 8525
AQA 8525 的 Paper 1 叫 Computational thinking and programming skills，整卷 2 小时、90 分，占 GCSE 的一半。规格说明写得直白：这一卷考计算思维、代码追踪、问题求解，以及有效算法的设计，还有代码的编写、测试与改进。题目来源限定在考纲的 3.1 Fundamentals of algorithms 和 3.2 Programming 两节。考生要选一种语言作答，AQA 提供 C#、Python 3 和 VB.NET 三个版本，题目会明确说明何时必须用指定语言写代码。代码写在纸上，没有编辑器提示语法，也没有运行按钮告诉你结果对不对。卷面按题目要求的形式作答，伪代码、程序代码还是流程图，题目说哪个就写哪个。阅卷时看不到代码是否跑通，能拿分的只有结构、语法与逻辑本身。常见的错误是把缩进和冒号当成小事，写在纸上时这些细节一个都不能省。
AQA keeps both papers on paper. Paper 1 lasts two hours and carries 90 marks, half the qualification, and candidates answer in one of three programming languages. Because nothing can be run, a misplaced colon or a missing indent costs marks that a compiler would have flagged. Many students lose marks here by answering in pseudocode when the question demanded program code, which is a habit worth breaking early.
三、Pearson Edexcel 1CP2 的 Paper 2 是真机考 / Pearson's on-screen practical
Pearson Edexcel 的 GCSE 计算机科学（1CP2，2020 年首次教学，2022 年首次考试）是三个局里唯一把编程放进机考的。Paper 2 叫 Application of Computational Thinking，2 小时、75 分，占一半成绩。官方说明这是一次实操性的上机考试，考生要设计、编写、测试并改进程序来解决问题。考场提供代码文件、纸质题目和 Programming Language Subset（编程语言子集），答案用 Python 3 在机器上完成，编辑器可以用自己熟悉的 IDE。全卷六道必答题，每道都要求学生动手写出能跑的程序，而不是描述步骤。两卷合计 3 小时 30 分、150 分，笔试与实操各一半。很多学生以为机考等于更容易，实际上自动补全和语法提示会掩盖理解上的空白，一旦题目要求解释为什么这样写，答案就露馅了。常见的错误是写完就跑，不回头看边界输入。
Pearson is the odd one out because it assesses programming on screen. Paper 2 runs for two hours, carries 75 marks and accounts for half the grade, and the board describes it as a practical on-screen examination in which students design, write, test and refine programs. Candidates work in Python 3 with the IDE they are used to, helped by a Programming Language Subset insert that limits what they are expected to know. An auto-complete prompt can hide a shaky understanding, so the written explanations attached to each task still carry real weight.
四、Cambridge OCR J277 的 ERL 书面卷 / OCR's reference language
Cambridge OCR 的 J277（2022 年首次考试）把编程技能留在书面卷里，靠一套公开的 Exam Reference Language（考试参考语言，简称 ERL）来考。Component 01 Computer systems 和 Component 02 Computational thinking, algorithms and programming 各 1 小时 30 分、80 分，各占一半。J277 规格说明把 Practical Programming skills 的评估明确归到 Component 02，考生要读、要追踪、也要写出 ERL 代码。这套语言是中性语法，学生课上用 Python、C# 还是 Visual Basic 都不影响评分。规格里没有类与对象，也没有 try 与 catch，异常处理在 GCSE 层面用输入验证和防御性设计代替。编程项目在课程中要做，但不计入最终成绩，这一点常被误传。写 ERL 的关键是记牢它和 Python 的差别，比如循环与输出语句的写法不同。
OCR takes the opposite route and keeps programming inside a written paper. Components 01 and 02 are each 1 hour 30 minutes, 80 marks and half the qualification, and the specification places practical programming skills assessment in Component 02. Candidates read, trace and write code in the Exam Reference Language, a neutral syntax that allows students taught in Python, C# or Visual Basic to be marked against one standard. There is no object orientation and no exception handling at this level, and the programming project carried out in lessons does not contribute to the grade.
五、三个考试局的评估方式对照 / Comparing the three routes
编程卷对照一览
AQA 8525：Paper 1 笔试，2 小时，90 分，语言在 C#、Python 3、VB.NET 中选；Paper 2 笔试，1 小时 45 分，90 分。
Pearson Edexcel 1CP2：Paper 1 笔试，1 小时 30 分，75 分；Paper 2 机考，2 小时，75 分，用 Python 3 上机作答。
Cambridge OCR J277：Component 01 笔试，1 小时 30 分，80 分；Component 02 笔试，1 小时 30 分，80 分，用 ERL 作答；编程项目不计分。
把三份规格说明放在一起，差别一眼就能看清。AQA 是全笔试路线，编程卷 2 小时、90 分；Pearson Edexcel 是笔试加机考，编程卷 2 小时、75 分；OCR 也是两卷笔试，编程卷 1 小时 30 分、80 分。三种安排对同一份考纲内容的权重几乎相同，区别在考试媒介和答题语言。语言选择上，AQA 给三种，Pearson Edexcel 只认 Python 3，OCR 用自家中性语法。同一道算法题，三个局要求的输出形式并不一样，题目说写流程图就不能交伪代码。选科时如果学校只提供其中一个考试局，复习重点就随之固定下来。转学的学生要特别注意，换局等于换答题语言，原来的肌肉记忆可能用不上。
Put the three specifications side by side and the medium is the clearest difference. AQA keeps both papers on paper, Pearson Edexcel examines programming on screen in Python 3, and OCR uses its own neutral syntax inside a written paper. The content they test overlaps heavily, so a student moving between boards mainly has to relearn the answer format rather than the subject itself.
六、机考与笔试实际考的能力差 / What each medium really tests
纸质卷和上机卷考的核心能力有重叠，也有明显缺口。两者都要求读代码、追踪变量、找出逻辑错误，也都要求写出完整程序。差别出现在反馈与调试上。上机考试可以运行程序，边界数据对不对当场就知道；笔试只能靠心算和纸面追踪，一步错就整段错。上机卷容易高估学生的真实水平，因为编辑器会补全函数名和括号；笔试卷则容易低估，因为排版和时间压力吃掉了部分能力。备考时可以把两种练法都用上：平时在电脑上写，考前两周改回纸笔，练手写缩进和语法记忆。阅卷时看到的错误集中在变量未初始化、循环边界和测试数据不全这几类。
The two media overlap in what they reward but not in what they punish. Reading code, tracing values and spotting logic errors matter in both, yet an on-screen paper gives instant feedback when a program is run, while a written paper leaves every trace to the candidate's own working. Practising only on screen tends to flatter a student, because editors supply function names and brackets that no exam desk will provide. Two weeks before the exam, switching back to paper is a cheap way to find out what has really stuck.
七、编程项目为什么不计入成绩 / Why the project carries no marks
三个局的 GCSE 计算机科学都没有把编程项目算进总分，这一点经常被家长和学生搞混。AQA 8525 的规格写得很清楚，两卷之外没有课程作业，但学校仍必须给学生足够的编程实践，用来支撑笔试里的编程题。OCR J277 的编程项目要在课程中完成，属于不计分的学习环节。Pearson Edexcel 1CP2 干脆把编程搬进机考卷，用一次 2 小时的实操代替长期项目。这样安排的原因不复杂：项目作业的评分一致性一直是个难题，受控条件下的一次考试更容易控制可比性。很多学生以为项目做得好就能补回笔试丢的分，这个误解代价不小。分数只来自那两场考试，这一点在三个局里没有例外。
None of the three specifications counts a programming project towards the final grade, which surprises many families. AQA states plainly that there is no coursework, though centres must still give students genuine programming practice to support the written questions. OCR keeps a project in the course without awarding marks for it, and Pearson replaces long-term project work with a single practical on-screen paper. The reasoning is consistency: marking a project fairly across thousands of centres is far harder than marking one script sat under controlled conditions.
八、Ofqual 机考监管咨询的控入口径 / Ofqual's controlled approach
Ofqual 在 2025 年 12 月 11 日启动了 12 周的机考监管咨询，到 2026 年 3 月 5 日截止，重点是怎么把机考管起来。咨询文件里的第一条提案是把默认值改掉：以后 GCSE、AS 和 A level 原则上不允许机考，只在少数情况下例外。例外之一是每个考试局最多可以提交两个新的机考规格，并且要经过 Ofqual 的认证。另一个限制是报考人数超过 10 万的科目现阶段不准做机考，数学、英语和科学这些大科目因此被挡在门外。考生不能用自己的手机、笔记本或平板，设备由学校提供。机考版与纸笔版是两套独立的资格，题目也不同，不是同一张卷子的两种作答方式。Ofqual 表示会在 2026 年公布回应摘要与后续安排。这条咨询的走向会影响未来几年编程考试待在纸上还是回到屏幕上，对在读学生来说，眼下按现行规格准备就是最稳的做法。
Ofqual ran a 12-week consultation on on-screen assessment from 11 December 2025 to 5 March 2026, and its first proposal would flip the current default so that on-screen assessment is no longer generally permitted in GCSEs, AS and A levels. Two limited openings are proposed: each board could submit up to two new on-screen specifications for accreditation, and subjects with more than 100,000 entries would stay out of scope for now. On-screen and paper versions would be separate qualifications with different questions, and students would not be allowed to use their own devices. Ofqual has said it will publish a summary of responses and next steps during 2026, so the detail still has to be settled.
九、计算机科学为何留在机考豁免名单 / Why computer science is exempt
咨询提案留了一条老路：当机考对评估有效性很重要、或者本来就是考核方式的一部分时，它仍然被允许。Ofqual 列出的现成例子包括计算机科学、英国手语和音乐技术。手语与音乐技术能留下，是因为这些技能在纸面上根本没法完整呈现。计算机科学的理由类似，代码写出来跑不跑得通，本身就是评估内容。已经上市的机考部分不受新规影响，Ofqual 的研究文档点名了 Pearson 的 GCSE 计算机科学、WJEC Eduqas 的 GCSE 计算机科学、GCSE 地质学与食品准备营养学，以及 AQA 的 A level 计算机科学。把编程放进真实开发环境来考，是这个学科公认的有效做法，出题人也一直用它来区分背答案和真会写程序的学生。这意味着短期内编程评估的分水岭不会从机考与笔试之间消失。对考生来说，选局之前确认自己考的那张卷子是上机还是纸笔，比纠结改革时间表实惠得多。
The consultation carves out one route that already exists. Where on-screen delivery is important for validity, or integral to the assessment approach, it stays permitted, and Ofqual names computer science alongside British Sign Language and music technology. Existing on-screen components are unaffected, including Pearson's GCSE computer science, WJEC Eduqas GCSE computer science, GCSE geology and food preparation and nutrition, and AQA A level computer science. Running code in a real development environment is a defensible way to assess programming, because it separates students who can write working logic from those who have memorised an answer.
十、报考人数下降 7.2% 与选科判断 / Falling entries and choosing the subject
Ofqual 在 2026 年 8 月 20 日发布的 GCSE 成绩概览里，把 computing 列为报考人数变化显著的科目之一，降幅 7.2%。同一年英格兰共发出 5,720,045 份 GCSE 成绩，4 分及以上比例为 67.0%。这些数字本身不说明编程评估变难了，但它提示选科格局正在变化。把报考趋势当成选科理由并不划算，兴趣与后续课程衔接才是。对打算选计算机科学的学生，更实际的判断是看目标高中或第六学级学院用哪个考试局，再对着那一份规格说明准备。大学申请阶段看不到考试局信息，招生官看到的是分数。技术雇主更在意你会不会写能跑的程序，这项能力在机考卷上练得更直接，在笔试卷上要靠纸面追踪来替代。
Ofqual's GCSE results summary, published on 20 August 2026, lists computing among the subjects with a notable change in entry, down 7.2% on the previous year, against 5,720,045 GCSE results issued in England and 67.0% of them at grade 4 or above. The trend says nothing about whether programming has become harder to assess, but it does say the cohort is shifting. For a student choosing the subject, the useful question is which board the school enters candidates for, followed by a careful read of that specification. Employers care far more about whether a program runs than about which board awarded the grade.
十一、测试数据三类与阅卷扣分点 / Test data and where marks go
三份规格说明都把测试数据分成普通、边界与错误三类，这部分最容易带着答案进考场。普通数据验证程序在正常输入下能跑，边界数据测的是循环起点终点与数组首尾这些位置，错误数据考的是输入验证会不会拦住不合理的内容。写代码前先把测试表列出来，通常会顺手发现自己的循环条件有问题。常见的错误是只测一组漂亮数据就交卷，剩下的分数全丢在边界上。OCR 的 ERL 没有异常处理语句，遇到非法输入要靠验证分支解决，这一点和很多学生课上学到的写法不同。AQA 的编程卷里，题目会明确要求写出测试结果，而不只是写代码。Pearson 的机考卷则要求设计、编写、测试与改进四个动作都留下痕迹。
All three specifications ask for normal, boundary and erroneous test data, and that vocabulary is worth memorising because it appears in mark schemes across boards. Boundary cases are where students quietly lose marks: the first and last pass of a loop, an empty array, a value sitting exactly on a threshold. A test table written before the code usually exposes a faulty loop condition that would otherwise survive to the exam. OCR candidates cannot reach for exception handling, so invalid input has to be caught by validation branches instead.
戴睿（dere）Neo优选系列 创意双触摸屏高性能笔记本电脑大学生AI学习商务办公设计炒股娱乐手提轻薄本32/1TB¥4586.0📱 去京东购买京东精选
需要一对一辅导？Need One-on-One Tutoring?
A-Level / GCSE / IB / KS2-KS3 全科辅导，覆盖 AQA、OCR、Pearson Edexcel 考试局
16621398022 同微信
Visit aleveler.com
关注公众号 tutorhao 获取更多学习资源
或截图后用「扫一扫 → 相册」选择本图