The Deep Dive, a full sample

Amara is in Year 12, predicted A* Maths, A Further Maths and A Physics, with a fourth A-level alongside. GCSEs are an 8 in Maths and a 7 in English, which are above what every course here asks. She wants CS in a big city - more excited by AI and building things than pure theory. She's considering UCL, Warwick, Bristol, Birmingham and Sheffield.

The grades will get Amara into a top course. Two reaches leave her with one safety. It asks AAA, a grade below her prediction. The list needs one more course that still works at AAB. The application itself needs to be specific to her.

Keep the Maths at A*, add a safety that asks less than AAA, keep UCL on the list and finish the project over the summer.

You told me Amara wants a big city, that she's more interested in how software gets built than theory for its own sake and that you'd consider a joint course if the computing stays central. You also asked me to look at Oxford and Cambridge, so both are on the list. Your worry is what happens if the grades come in below the prediction. That worry is exactly right and it's the first thing I've fixed.

These are real UCAS course figures from 2023 to 2025. Achieved means the grades students arrived with. Offer rate is the share of applicants who got an offer, so it's competition and not a personal chance. Pay is the median for that subject at that university five years after graduating, from the 2023/24 tax year, so every computing course at one university shares the figure. Confirm the entry requirements on each course page for 2027 entry. No outcome is guaranteed.

The five are all famous names. Her Maths, her Further Maths and the applied-computing interest are what CS departments outside the top few look for. UCL makes offers to 10% of applicants. The courses in the middle of this list make offers to between 60% and 80%. So the first thing I'd fix is the middle of the list. Make that strong and nothing on the form depends on UCL.

UCL - keep it, don't rely on it. It has an offer rate of 10%.

Warwick asks A*A*A, a grade above her prediction, so it's a reach on the grades. 70% of applicants get an offer though, against 10% at UCL. If the September mocks come in at A*, Warwick is where I'd put the firm.

Bristol asks A*AA, exactly her prediction. It takes 60%.

Birmingham asks A*AA too. It takes 71% and students have got in there with ABB.

Sheffield asks AAA. That's a grade below her prediction. It takes 80%.

Oxford and Cambridge are here because you asked for them on the call. The rest of the extra courses are the ones this list is missing - more genuine safeties, the strong departments outside the famous names and the big-city options Amara asked about.

The joint courses are worth the look. Oxford has an offer rate of 10% for straight CS and 10% for CS and Philosophy, so that one is no easier. Sheffield asks BBB for its AI course against AAA for straight Computer Science.

At AAB two courses still ask no more than she has, Liverpool and Newcastle. Fourteen have taken students with AAB.

If anything slips, it can't be Maths. Her A* is in Maths and every course on this list asks for it.

There are two ways to build your five courses. Which one depends on what the Maths does between now and results day.

One list keeps both reaches. UCL and Warwick as the reaches, Bristol and Birmingham as the targets, Sheffield as the safety.

The other gives up a reach for a second safety. UCL as the only reach, Bristol and Nottingham as the targets, Sheffield and Glasgow underneath. Both of those ask AAA, a grade below her prediction.

Decide after the September mocks. Maths still at A*, take the first. Maths slipping, take the second. Either way the firm and the insurance need to ask for different grades.

Five sensible choices are either two reaches, two targets and one safety or one reach, two targets and two safeties, depending on how much risk you want. Amara's five are the first of those two, so there's only one course below her. Sheffield asks AAA and has an offer rate of 80%. Students have got in there with ABB.

The longer list gives her three ways to add a second safety. Glasgow asks AAA, the same as Sheffield. Nottingham asks A*AA but makes offers to 82% of applicants. Sheffield also runs Computer Science with Artificial Intelligence at BBB, three grades lower than its straight course.

Any one of those underneath means she keeps UCL and still has something that works at AAB.

If the firm and the insurance ask for the same grades, the insurance isn't doing its job.

You'll decide which course goes where months from now, with real offers in front of you.

Admissions tests. Some top CS courses use a test or an interview. Check each course for 2027 entry and register early.

GCSE rules. The 8 and 7 clear every course here, no issue.

Subject rules. Maths is required everywhere and Further Maths helps at the most competitive courses. Amara has both.

The fourth A-level. No course here asks for it.

One statement has to cover straight CS everywhere and possibly Philosophy and CS at UCL. What works for both is how systems make decisions. Write about the thing Amara builds this summer, what she read because of it and the question it left her with. Name a module or research group at one or two of the five courses. Every CS applicant lists languages and frameworks, so use the space to say what she made of them.

Here's what to do, from now to the day she applies.

Book the open days now and put in two work-experience applications.

Over the holidays, work on the project and keep the Maths ticking over. Nothing heavier.

By the mocks, pick between the two lists and register for the tests.

In the term before the deadline, draft the statement, settle the five courses and submit.

When the offers come in, choose the firm and insurance at least a grade apart.

The dates that can't move.

15 October. Oxbridge and medicine close. Nothing on this list needs it but it's the date to know.

13 January. The equal-consideration deadline for everything else on her list.

Results day. She has already chosen the firm and insurance, at least a grade apart.

Where a test has a registration window, the booking-close date matters more than the sitting date. Put the closing date in the calendar.

Keep UCL and Warwick. Warwick is the more likely of the two reaches at 70%. Add one more course that asks below AAA. Finish one piece of software over the summer.

Which of the two lists, what goes at the bottom of it, the final five courses. We go through them on the second call. I'll send a revised report within seven days of it.

I put three questions to the admissions offices for you. Their answers, who gave them and the date.

These are the questions the tables can't answer. Whether the Further Maths counts toward the offer at the two courses that ask for it. Whether the BTEC route is accepted alongside two A levels at the one course whose page is silent on it. Whether the placement year affects the offer at the two that run one.

This sample shows the questions rather than the answers. Amara is invented, so no university has been asked about her.

One figure isn't enough to read a course by. A one-year figure says almost nothing, because most of these graduates are still settling.

The five-year figures spread much wider than the three-year ones. Warwick climbs from £54,000 to £71,400 and Bristol from £53,800 to £70,100, while Manchester moves from £55,300 to £57,900 and Leeds from £47,000 to £49,800. Warwick and Manchester are within £1,300 of each other at three years. At five they are £13,500 apart.

The middle half is what the middle 50% of graduates earn. At Cambridge it runs from £63,300 to £198,700, which is a wider spread than the whole of the rest of this table.

Oxford publishes a one-year figure for computing only, on twenty graduates, so nothing is shown here.

One line per subject she takes, from the DfE value-added figures for her own sixth form.

Value added is what her school does with students who arrive with her GCSEs. Above zero means they leave with better A levels than the same students got elsewhere.

DfE doesn't publish a figure where the year group is too small. Further Maths is missing at about half of schools.

Where a number is printed it's often too close to average to mean much, so I only call out the subjects where the school is ahead or behind.

This sample has no figures. Amara is invented, so she has no school.

Written now, in August of Year 12, so there's a plan ready before results morning.

Missing the grades by a little doesn't always cost her the place. Clearing entrants get in about three grade points lower than main-scheme entrants, which is roughly a grade on each of three A levels. The courses with places in August are often not the ones families expect, so this plan uses the longer list.

From 8am. Have the UCAS ID and the list of twenty to hand. If the grades come in, nothing to do. If they're one under, ring the firm before doing anything else, because a near miss is often confirmed anyway.

If it goes further than that. She still meets Liverpool and Newcastle at AAB. Students have got into Southampton, Leeds and Sheffield with AAB. All three make offers to 75% or more.

Nothing here is a promise of a place. It's a list of what to do, in order.

Ten things to build toward, each one named, with when to do it. This list is built around what Amara is interested in, so another student with the same grades could get a different ten.

Start with the coding. Kaggle Learn is free and comes in bite-size courses. It's the fastest way to get Python solid enough to build something with.

Autumn of Year 12. Enter the UKMT Senior Maths Challenge through school. It backs the Maths A* and the CS application at once.

Autumn, once the Python holds up. The Kaggle Titanic competition is made for first-timers, real data and a gentle start. Under 18s need a parent to approve the account.

Winter. The British Informatics Olympiad is entered through school and it's properly hard. Sit it after the first two, not before.

Spring. DeepLearning.AI is free to audit and it's where the theory behind the project comes from.

Spring. Apply to two work-experience programmes. Uptree and Springpod both list ones teens apply to directly and several of the big employers run their own.

The summer before she applies. Finish one project and put it on GitHub. Software that solves a problem she actually has. It's the strongest single signal she can send and it's what the statement is built on.

The same summer. One book and one lecture series on how algorithms decide things, then write about them.

A stretch, only once she's genuinely strong. The UK AI Olympiad, after the project and the Informatics Olympiad rather than instead of them.

The timings here are when to do each thing, not the published deadlines. Anything that has a closing date is given as at 30 August 2026, so check it before relying on it.