At the chalk face: It's not in the mark scheme

At the chalk face: It's not in the mark scheme

Ever marked a mock and heard, “But it’s not in the mark scheme…”? 🤔
In this episode of At the Chalkface, Craig and Dave tackle one of the biggest misconceptions in Computer Science assessment — and it might just change how you mark (and teach) forever.

Drawing on real examiner experience, they unpack the truth behind mark schemes, including:

Why mark schemes are not exhaustive (yes, really!)
How students can still gain marks for valid answers not listed
The balance between teaching the spec vs the mark scheme
Why “correct” in Computer Science isn’t always as black and white as you’d hope
And how algorithms like quicksort highlight just how many right answers can exist

💡 Key takeaway:
It’s not about what’s in the mark scheme… It’s about what’s correct.

Whether you’re a specialist or teaching out of your comfort zone, this episode will give you the clarity and confidence to approach marking — and student queries — in a whole new way.

👉 If you’ve ever second-guessed a mark or been challenged by a student, this one’s for you.

Get our book: Essential algorithms and data structures: Suitable for all exam boards https://amzn.eu/d/0c50n9rE 

🔔 Don’t forget to like, subscribe, and explore more support for teaching Computer Science with Craig’n’Dave.

#ComputerScience #GCSEComputing #ALevelComputing #TeachingTips #TeacherLife #EdTech #CraigAndDave

Transcript

SPEAKER_01 0:00

It's not in the Mark scheme. Does that mean students are awarded a mark? Yes.

SPEAKER_00 0:10

Hello, it's Craig and Dave here, and welcome back to another At the Chalkface episode. And today we're talking about mark schemes. And more specifically, I mean we've titled the episode, it's not in the mark scheme. It's like uh, yeah, there you are, you're you're doing a mock with your students, and now you're marking it together, you've swapped papers, and it's like, oh, I've got this. Well, it's not in the mark scheme. Is it worth a mark? Well, the answer might be slightly more nuanced than you think. And do not feel bad about this as a teacher if you come away from this thinking, oh blamey, that's not how I've been doing it. It wasn't how I did it at the start. And uh we'll unpack some of this. So um, yeah, there's there's some interesting stuff to talk about here. But Dave, shall I shall I start with um my limited, but uh my my actual experience as an exam on this? Because this was the moment when I actually learned for the first time how different this was to how I thought. So maybe I should open with that one go from there.

SPEAKER_01 1:07

Yeah, absolutely. You've been an examiner, Craig, so spill the beans, tell us what it's all about.

SPEAKER_00 1:13

Okay, now our first caveat, this was a number of years ago, and obviously advice always um evolves. Uh, but I was an examiner for the OCR uh GCSE exam board for another year, and obviously you go to training sessions, and uh it was the chief examiner uh at the time who sat there leading this session with all those keen teachers, and he started up by saying, So, first thing I want to let you all know as future examiners and markers is there are two mark schemes. There's the one we publish and we give out to teachers we put on the website once the exam series is done, and there's the one we use. You what? Do you want to say that again? Then of course he unpacked that, which I will do briefly now, and of course it made perfect sense. He said, When we write a marked scheme, we look at the specification content and we write the sort of answers we'd expect students to provide based on their depth of the level of study and what's in the specification. However, our mantra as OCR and your mantra examiners is we award students for getting it right, not for getting what's in the MART scheme. We do not penalize students for having a correct answer that goes beyond the breadth of study, maybe to A level, or beyond the list of examples that we specifically state we will ask about. What that means is the number of possible right answers can far exceed those we list in the Mark scheme. We can never publicize all the right possible answers we give a Mark for because the MARC schemes will be hundreds of pages long. And even the Mark scheme you'll get as an examiner won't list every possible answer. You might come across something a bit left of field, a bit nuanced, you've got a super keen student, and you'll be there going, oh, that's where you need to escalate it to your team leader. This is all a long way of saying that before then I was probably being too strict with my students. Well, it's not in the Mark scheme, you won't get the mark. Now, I don't know what the rules are for AQA and Excel, but I have a funny feeling they are very broadly similar.

SPEAKER_01 3:24

Yeah, they don't want to be able to do that.

SPEAKER_00 3:25

We want to be crediting students for correct answers. We don't want to penalize them because they gave a correct answer beyond the spec that wasn't in the mark scheme. So I think that's the first big takeaway. Just because it's not in the Mark scheme doesn't mean it's right. Yeah, Mark schemes, well, the key bit here is Mark schemes aren't exhaustive.

SPEAKER_01 3:46

Yeah, it's worth saying though, that um the Mark schemes aren't just cobbled together, right? I mean, what happens is uh a sort of I'm gonna use the word rough here, but a rough mark scheme is created for the question with what the examiner, the person that wrote the question maybe thinks would be um appropriate answers given the uh depth and breadth of the level of study, as you've already said, Craig. And after that, then there's a meeting uh where people sort of get together and sort of thrash out um some of these possible answers in order to expand the Mart scheme a little bit and try and capture the range of possible answers that students can give, because you don't really want to be in a situation where there is actually an obvious answer to a question, but the examiner who wrote the question didn't think about that at the time and then it didn't make it into the Mart scheme. So it's worth saying that the Mart schemes do contain the breadth of answers that you would anticipate students using. So I definitely wouldn't be advising the students to deliberately find um, you know, an edge case to deliberately find something that is a bit obscure, that is, you know, factually correct, but you're not going to find it unless you do some deep research. It's it's not worth it, you know. Teach what's in the mark scheme because examiners know that these uh these MARC schemes are in the public domain. And so, in a way, it can uh very much focus your teaching if you're a little bit unsure on, you know, well, what are the actual things that I should be teaching in this section of the specification? Yes, there are delivery guides and other bits and pieces um of support material out there, but the MARC schemes are a great way of focusing your teaching and thinking, right, these are the important things um that I should teach here. So it is it is good for that. Um, so I wouldn't be advising the students to be deliberately looking for something that's factually correct, just to kind of argue the point. Because it's worth saying, Craig, that examiners are just ordinary teachers. At the end of the day, let me tell you some some truth if if there's members of the public watching this. The truth is examiners are just teachers that want to earn some extra money in the for the summer holiday. That that is the truth of it, okay? And no teacher has the monopoly of knowledge, um, they won't know everything about every discipline of computer science. And it it definitely is the case that they will be using the MARC scheme as the guide. And if a student writes an answer that might be factually correct but is a little bit odd, they do risk getting that marked wrong, and that could potentially affect their grade, and that's why there's an appeals process, and that's why um you know you you can sometimes get these grades changed as a result of that. But it's largely about administrative errors these days, if I'm honest. But Craig, I've I've talked a lot there about uh uh about all that. Is that fair?

SPEAKER_00 7:06

Yeah, no, no, that is fair. Um, and because I I I remember again the chief Red Two training sessions, the chief examiner saying, you know, you won't know all the correct answers, and it will get very easy once you've been marking for several hours. Because you, by the way, again, you tend to mark the same question multiple times for consistency, not go through a paper. And when you've marked the same question 30, 40 times and your online system is showing you the next one, um, dare I say it, you can almost become a bit blasé. Now, I don't want you to think marking is lazy and exam is not as robust processes, and your team leader can view you and watch what you're doing, but you know, it it becomes easier. Oh no, that's the wrong answer. Whereas maybe half an hour later I might have paused for thought or thought, oh, I escalate that to my team leader. Dave's absolutely right, you know, teach, don't teach to the Mart scheme, but teach to the spec and encourage students to put the the obvious answers that are there. You know, the more obscure the answer, the more you run the risk of it not being accredited a mark when it actually is value of one. Now, you actually see this to a certain extent in the MART scheme anyway. Obviously, we've got certain questions, you know, like the binary questions are obvious one, you know, state uh also state the register that holds the address of the next instruction. Things like that. There is an answer, there is an answer, it's right or wrong. The the the beauty of marking a maths paper. But um unfortunately since you've mentioned register, I I have to take a pause there.

SPEAKER_01 8:36

Go on, Dave, interrupt. You know where I'm going. It is a specific example, right? So if if the question was um state the name of the register that holds data coming from the memory to the processor, one mark. Right in the OCR specification, you would anticipate the students writing the memory data register, the MDR. Oh, and by the way, I would write memory data register, not just MDR. You probably get away with MDR, but yeah, you know. Um, and if you can't remember it as the memory data register, but you can remember it as MDR, then write MDR, right? Anyway, I've gone through a sidetrack here. The the point I'm trying to make is it might only say memory data register in the Mart scheme because that's what it says in the specification. However, with that example, if you said memory buffer register, it is definitely worth a mark.

SPEAKER_00 9:44

Yeah, a hundred percent. Uh-huh. And just to uh emphasize that point, in the OCR spec, it is called the memory data register, in the AQA spec, it's called the memory buffer register, and that's probably the only mark in each of those MART schemes, but both are interchangeable, both are equally valid, and both you want a mark. But I would encourage students to put the one in the Mart scheme. But where I was going before Dave rudely interrupted with me, which he so often does, um we we see this in in the MART scheme in questions where they ask you to code, and they will actually write it. They'll say, Oh, well, here's a here's an example of the of the pseudocode which we we expect them to write for this for this sort or this algorithm, and it will say, but except other um options which which meet these points. And algorithms is the great one. I've seen so many debates on forums, and uh I know Dave wants to talk about a particular algorithm in a minute, so I won't take his thunder on on that, but I see so many fights, almost fights on forums over the years. Like, that's not how you do uh a stack, for example. The pointer should point to the top item of the stack. No, it should point to the next available space above the stack. It doesn't matter as long as you're consistent. And actually, in my algorithm videos, uh I say if it looks like an algorithm and it smells like an algorithm and it tastes like an algorithm, it's probably the algorithm. And what I mean by that is when you're physically writing lines of code to implement a stack data structure, what actually makes that a stack? Well the last item in should go at the top. And when we pop it, it should be removed. Can you pop items from the top? Can you push items from the top? It's the core functionality. Should I be able to look at an item in the middle of a stack? If I implement a procedure to peek into a stack, sure. Should I be able to reverse the contents of a stack? Well, that's not traditionally what happens with a stack, but if you added an extra routine to reverse the item of a stack, it doesn't take away from it not being a stack as long as you've implemented the main features. The implementations can be vast. And that particular thing I mentioned there was a classic. The pointer pointing to the top item of the stack or the first space, it doesn't matter as long as you're consistent. If it looks like the algorithm, it smells like the algorithm, it is. And um, I know you want to talk about a particular favourite algorithm of yours that really exemplifies this, Dave, and gets to our point where just because it's not in the Mart scheme doesn't mean it isn't right. It's our that's our friend the quick sort, isn't it?

SPEAKER_01 12:23

Yeah, it it is. Uh before I talk about the quick sort, it I do, if you're a non-specialist computing teacher, I feel your pain because you would expect there to be right answers. You know, you expect to look at the Mart scheme and say, that is the answer. And when students then challenge you, you need to you want to have the confidence to be able to say, no, what it says in the MARC scheme is correct. But unfortunately, it is the nature of the subject we teach. Yeah. That in computing, there are there is more than one right answer. I mean, sure, some algorithms are more efficient either in the amount of memory that they use or the amount of processing time that they take. But if it works, it works to a certain extent, particularly at GCSE and A level. And as Craig said, there's all sorts of additional methods that you might apply to some of these uh algorithms and data structures to increase the functionality of them. Are they in the specification? No. Are they expected knowledge? No. Could they be done? Yes. So it's it is really, really tricky for the non-specialists that are grappling with subject knowledge and thinking, I don't know what's right and wrong. Um that kind of comes with experience, unfortunately. And as Craig said, what you're really looking for are the features of whatever it is rather than specific code. And it's worth saying at this point, please, please do not teach the students to memorize code blocks. Yeah, for example, you know, if they are asked for um the code for a bubble sort, for example, don't memorize the code, memorize the procedure. What what actually happens? What are the steps? What does a bubble sort do?

SPEAKER_00 14:21

How does it work? What is the essence? What is the essence of that algorithm? What are the key features that make that algorithm? That is much more because there'll be the things in the Mart scheme that they say, do they have a line of code that does this? Do they have a line of I'm making these up now? Do I have a line of code that finds a pivot, increments a pointer, does this? And they're the key features that make that algorithm that algorithm. And then they'll say, and here's one example of how that might be implemented. So come on then, tell us about your favorite algorithm, your quick sort, and good old is it Tony, Tony Hoare? Is that Tony Hoare?

SPEAKER_01 14:55

And and we've got sad news. I mean, you may have heard this already, but um unfortunately, uh Tony Hoare has sadly passed away uh peacefully, and uh he was the inventor of the uh quick sort, and uh not just the quicksort, I mean he has given so much to the field of uh computing uh that uh you know I'm not doing him um any justice. But um yeah, Professor Sir Charles Haare, uh known as Tony because his middle one of his middle names was Anthony, um, has unfortunately he's unfortunately no longer with us. But that's really um what kind of made me think about doing this episode because I wanted to acknowledge and reflect the contribution that Tony has made to our subject um in a respectful way, um, but also in a way that sort of resonates with some of the problems that teachers and students are facing in the classroom today at the same time. So, you know, hopefully this is um uh a way that we can uh sort of show our respect um to Tony. But anyway, the um the quick sort, yes. Um so Tony invented the the quicksort, and there's um a few a few stories around sort of how that came about and how he was um challenged to uh write a shell sort by his boss and said, Well, I've got a better way of doing that, and uh he won a sixpence as a result of proving that the quicksort is the best sorting um algorithm. It's worth saying, like all algorithms, uh the quick sort is generally the quickest sorting algorithm, but not necessarily in all cases. And again, this is where you get situations in the classroom where, you know, uh a student might say, What's the best sorting algorithm? And the teacher says, the quick sort. And then someone else says, No, the bubble sort is the quickest. And you're like, no, the bubble sort definitely isn't the quickest. But actually, the bubble sort can outperform the quick sort in a certain situation. For example, if the list is already sorted. So it this is what we're grappling with in our subject, unfortunately. But yeah, the quick sort is really a family of algorithms. So although Tony Hoare created the original quick sort, it has had many variations and slight deviations over the years. I mean, one of the most famous ones that I won't get into is the Lamoto method. And uh some teachers would argue that is a slightly easier quick sort to teach. And so uh they will go down that road. And maybe that is the quick sort that they learned themselves. So it's the one they're familiar with. And you do get those situations then in the classroom where uh you teach a particular way of doing a quick sort, and the students will say, That's not how we learnt it in maths. Um, my teacher said the pivot never moves, it's nothing like the Mark scheme.

SPEAKER_00 18:07

You know, this textbook yeah, I've looked at this textbook and it's different in here. But that's the thing as well. You may be buying textbooks from other providers and they will show you the code for a quick sort. If it follows the steps that make it a quick sort, it's valid. It's probably almost definitely coded slightly differently to the one in a mark scheme and the one you learn and the one in maths.

SPEAKER_01 18:27

Yeah, absolutely. So, I mean, at the end of the day, if you are marking an answer about a quick sort, then you've got to look for the features that the student has correctly identified. Not the sort of almost individual lines of code or the particular approach because it might not be the one that you're familiar with. It might not be the one you taught. Okay. So at the end of the day, with a quick sort, you're looking for is there a pivot strategy? Is there a partitioning strategy? Is there some kind of divide and conquer strategy? Is there some kind of combination of the results? I mean, those are the critical aspects of the quick sort. That's what differentiates it really from the other sorting algorithms. This idea that you've got a pivot and the pivot causes partitions. If you've got that, then you've probably got a quick sort. But you know, what pivot is the student using? Are they using the first element in the list, the last element in the list, a middle item in the list, a random item in the list, a median of three, a median of five, Turkey's nine? There are loads of ways you can choose a pivot. Are some better than others? Yes, but it can be dependent on the data set that you're given. The point is, has the student got a pivot? And if they have, then it's probably working towards a quicksort. Have they got some way of partitioning the numbers? Because at the end of the day, the numbers uh to the left of the pivot should be lower, and the numbers to the right of the pivot should be higher. Is that always the case? I mean, could you have, for example, the higher numbers to the left of the pivot and the lower numbers to the right of the pivot? Yes, you absolutely could. It depends whether you want it in ascending or descending order. But the point we're making here is the pivot is the thing that decides whether numbers are lower or higher. We would call this a partitioning scheme. Does it does it work out where the pivot should be in the list? If it does, then you're on the way to uh a quick sort. You know, there's the classic um method, there's the Lilamoto method, there's the Bentley-MackElroy method, there's the dual pivot method that's used by Java. Uh, there are loads of partitioning schemes. So it's tricky looking for a very particular um algorithm. Then you've got this divide and conquer structure that we've talked about. And this is where we start disagreeing with our colleagues in maths, because in computer science, we're interested in the parallelism, the computational gains we can get through parallel processing. And therefore, to have a um a divide and conquer structure that enables us to give different cores different parts of the list to work on and combine them later is going to make it significantly more efficient. They're not going to teach you that in maths because it's it's just not necessary. So you've got things like tail recursion elimination cut off to an insertion sort, which means actually we'll look at the list and decide quicksort will be inefficient here. Let's just do an insertion sort instead. Um, you know, that that's a that's an option. And and all the implementations of uh sorting algorithms in programming languages today will be doing that. They'll be choosing the efficient the efficient algorithm. Have you got memory layout um optimizations? Have you got the Parallel variants that I just discussed. Have you got um uh intra sort hybrids? Have you got cash oblivious versions?

SPEAKER_00 22:07

That there are so many dividing conquerors. We've really gotten off the episode, and I've let Dave go down his love of that.

SPEAKER_01 22:15

They just have to because it it's just all I'm illustrating is that there are so many variants and variants and variants of quicksorts.

SPEAKER_00 22:23

There's not one quicksort.

SPEAKER_01 22:25

Yeah, you look at a mark scheme and you say, Oh, that's the that's a quick sort. And if a student does it in any other way, they're wrong. Yeah, no, they're they're they're not wrong. Um you know, my personal favorite is a uh a version of the uh whore method. Um the Hungarian dancers. Uh I love that. I mean you introduced that to me recently, yeah. Yeah, it is a it is a whore quicksort, it is a traditional quicksort with with one slight adaptation that the Hungarian dancers move the pivot, which you well, every time they do a swap, they move the pivot, which isn't the classic whore method, but it but it under the under the hood, it's it's a classic whore quicksort, really. Um but anyway, like Craig says, I've kind of gone off on one a bit, but I really wanted to illustrate the problems that you've got with MART schemes.

SPEAKER_00 23:17

Yeah, no, totally. I would be remiss at this point if I didn't do a quick plug for our data structures and algorithms book. Um we know an awful lot of teachers actually buy themselves one copy of this as a reference for themselves. Do not feel bad if this is an area that you struggle with. I struggle with it for a very long time. Uh if you're if you use our resource center and you have access, you've actually got a PDF copy for free, but you can buy a hard copy for about 15 quid off Amazon as well, and then annotate over it to your heart's delight. But we go through every single algorithm and data structure in all the GCSE and A levels, and we lay out step by step, starting with what are the key features of this algorithm, and I think that's the bit we were getting at. And then we provide pseudocode, we provide fully coded examples of the algorithm in multiple variants, we have visualizations and links off to our videos. So there is a lot of help and support there with that. But yes, to bring it all back, as we said, I think we've we've really nailed the point there. Just because it's not in the Mart scheme does not mean it's not credited and mark. We shouldn't be encouraging our students to deliberately go off piece and be as avant-garde as they like to show how clever they are. While we can reward them in the classroom for that, they are in danger of not being awarded a mark that they should do. So, you know, tell them to go for the safe mark if they can remember. But yes, it's not quite as simple as if it's not in the Mark scheme, you don't get a mark. Any closing comments from you, Dave, before we leave it there?

SPEAKER_01 24:47

No, I think that's it. Um sorry, non-specialists, I feel for you.

SPEAKER_00 24:55

But uh it's the nature of the beast. Perfect. Well, thank you very much, everyone, and we'll see you hopefully next week for another at The Chalk Face. Bye-bye for now.