Read on Twitter
I was going to wait till tomorrow but I can't.
The U grades (and other large drops) appear to be the result of a rounding issue when assigning ranked students to their grades. It'll be hard to explain via tweet but I'll try.
Please bear with me as I use a simplified example.
The U grades (and other large drops) appear to be the result of a rounding issue when assigning ranked students to their grades. It'll be hard to explain via tweet but I'll try.
Please bear with me as I use a simplified example.
Imagine a cohort of 20 students: each of these students (remember they're represent real people who have probably worked very hard alongside their teachers for 2 years - never forget that) is nominally worth 5% of the grade distribution however it falls.
Now the students, according to their teacher (excellent judgement - quoting DfE), have been assessed during this time and they've perhaps sat a mock or two and have done essays or end of unit tests and so the teacher submits a grade and a rank for each one.
The grade is a suggested grade for OfQual, but the rank is a distinct order and it ultimately will determine what grade each student will receive.
Now, coming up is the magic I will never know, because I don't have all the data, but OfQual have described it quite comprehensively.
Now, coming up is the magic I will never know, because I don't have all the data, but OfQual have described it quite comprehensively.
All those grades from these 20 students along with all the others from all the cohorts for that subject form a distribution of grades that OfQual can prod and poke at to see if it's a fair distribution.
They decided it wasn't, adjustments needed to be made.
They decided it wasn't, adjustments needed to be made.
Fair enough; it was to be expected. So they use a combination of the subject history at that school AND the prior attainment of the 2020 students to come up with a distribution of grades that the 20 students we imagined have to fit into.
For example:
A* 7%; A: 15%; B: 22%; C: 52%; D: 3%; E: 0.5%; U: 0.5%
I made these numbers up but let's pretend I ran them through "the algorithm".
This is what my 20 students have to fit into.
Notice I haven't told you their grades. It's now irrelevant. The rank is king.
A* 7%; A: 15%; B: 22%; C: 52%; D: 3%; E: 0.5%; U: 0.5%
I made these numbers up but let's pretend I ran them through "the algorithm".
This is what my 20 students have to fit into.
Notice I haven't told you their grades. It's now irrelevant. The rank is king.
Student number 1 gets an A*. Remember every student is worth 5% as I have 20.
Student number two would take us to 10% but wait I'm only allowed 7% so A grade for student 2 (Their CAG was A* - drat).
Students 3 and 4 take me to 20% so they get As (I'm allowed 22% to get A*/A)
Student number two would take us to 10% but wait I'm only allowed 7% so A grade for student 2 (Their CAG was A* - drat).
Students 3 and 4 take me to 20% so they get As (I'm allowed 22% to get A*/A)
My next 4 students all get B, we've used up 40% but I'm only allowed 44% A*/B. Sucks to be us. Next student (45%) gets a C.
I've got 55% worth of students (11) left but unfortunately only space for 10 in my Cs.
So I've one student left. We've used 96% of our grade allowance.
I've got 55% worth of students (11) left but unfortunately only space for 10 in my Cs.
So I've one student left. We've used 96% of our grade allowance.
Now here's the kicker.
Surely they'll get a D.
Nope, only 99% of my students can have a A*/D.
An E?
Fraid not that takes us to 99.5% not not a full 20 students worth.
But a U. I've got 100% now. So my student can go here.
And that is how rounding down gave C students U grades.
Surely they'll get a D.
Nope, only 99% of my students can have a A*/D.
An E?
Fraid not that takes us to 99.5% not not a full 20 students worth.
But a U. I've got 100% now. So my student can go here.
And that is how rounding down gave C students U grades.
And here's @FFTEduDatalab on a specific real example though not with the C to U bit. https://ffteducationdatalab.org.uk/2020/08/a-level-results-2020-how-have-grades-been-calculated/
The key thing though is if they'd rounded up not down, my 20 imaginary students would have had 2 A*s and no Us.
