TD
Key Takeaways
- Quantitative trading firms recruit very early, often opening applications more than a year before the role starts.
- The process is dominated by objective tests — mental arithmetic, probability, brainteasers and coding — rather than by fit and polish.
- Intakes are tiny, so rejection is the overwhelming base rate even for strong candidates.
- Compensation at entry is among the highest of any graduate role, and the work is narrower than most people expect.
- Preparation is unusually effective here because the assessed skills are explicit and trainable.
The three kinds of role
The terms get used loosely, and they are different jobs with different processes.
Quantitative trader. Makes decisions about risk and price, frequently in a fast, competitive environment. Assessed heavily on probability, mental arithmetic under pressure, and decision-making with incomplete information. Some roles are close to fully automated and supervisory; others are discretionary.
Quantitative researcher. Builds the models and signals. Assessed on statistics, machine learning, mathematical maturity and research judgement. Frequently requires a PhD or a very strong mathematical background, and the work is closer to research than to trading.
Quantitative developer or software engineer. Builds and maintains the systems — low-latency infrastructure, execution, data pipelines, research platforms. Assessed like an elite software engineering role, with more emphasis on performance and correctness. This is the largest of the three by headcount and the most accessible to a computer science graduate.
Which matters for your application. Candidates frequently apply to "quant roles" generically and prepare for the wrong assessment. A developer role will not ask you to price a dice game under time pressure; a trader role will not ask you to design a distributed cache.
Timing
Earlier than almost anything else on the campus calendar.
Internship applications frequently open in the spring or summer of the preceding year, more than twelve months before the internship starts. Some firms recruit on a rolling basis and close when they have filled.
Full-time hiring is heavily fed by internship conversion. At many firms the summer intern class is the primary source of graduate hires, which means the decisive application is made a year and a half before you would start.
Processes move very fast once started. Some firms compress the entire sequence into two or three weeks, and offers frequently arrive with short deadlines.
The practical consequence. If this is a target, you need to be ready in the spring of your penultimate year — not the autumn of your final one, by which point the main pipeline has closed. This is the single most common way capable candidates miss out.
What the process assesses
Unusually transparent, which is good news for anyone willing to prepare.
Online assessments. Typically a timed mixture of mental arithmetic, probability, pattern recognition and sometimes coding. Speed is the binding constraint — the mathematics is not advanced and the time allowed is short.
Mental arithmetic tests. A specific and trainable skill. Multiplying two-digit numbers, working with fractions and percentages, and estimating quickly, under a clock. Firms use these because they correlate with something they care about and because they cannot be faked.
Probability and expected value. Dice, cards, coins, urns, and games where you must decide whether to play and at what price. The mathematics is undergraduate probability; the difficulty is applying it fluently under pressure while explaining yourself.
Market-making games. You are asked to quote a two-sided price on something uncertain, then defend or adjust it as information arrives. This assesses risk sense, arithmetic and composure simultaneously, and it is the most distinctive part of trader interviews.
Brainteasers and estimation. Fewer than the reputation suggests, and still present. The point is your reasoning aloud, not the answer.
Coding. For developer and researcher roles, technical interviews resembling elite software interviews with more emphasis on efficiency and correctness. The standard preparation applies with the bar set higher.
Fit, briefly. Much less weight than in banking or consulting. Firms care that you are competitive, intellectually honest and can handle being wrong, and they assess that through how you behave during the technical parts rather than through separate questions.
Preparing properly
Train mental arithmetic separately and daily. Fifteen minutes a day for two months moves most people substantially. This is the highest-return preparation available because it is pure practice and it is directly tested.
Work through undergraduate probability properly. Expected value, conditional probability, distributions, and the standard puzzle patterns. Fluency matters more than breadth.
Practise thinking aloud. Every interview here assesses reasoning. Solving silently and stating an answer scores badly even when the answer is right.
Practise being wrong well. Interviewers will push back, sometimes when you are correct. Updating on a good argument scores well; capitulating to pressure when you are right scores badly; defending an indefensible position scores worst. This is deliberately tested and it is worth rehearsing.
Do market-making games with a partner. The format is unfamiliar and the first attempt is always poor. Doing five with a friend makes the real one your sixth.
Know your own resume technically. Any project you list will be probed to depth.
Being realistic
Intakes are tiny. Some firms hire a handful of graduate traders a year globally. The applicant pool is enormous and heavily self-selected toward strong mathematical backgrounds.
Rejection is the base rate, and it says very little. Excellent candidates are rejected routinely, frequently on a single timed test.
Which means: apply widely across the sector, prepare seriously, and do not organise your whole autumn around it. The placement resilience point applies with force — this is a sector where the arithmetic guarantees mostly rejections.
And have a genuine second track. Candidates who prepare exclusively for quant interviews and are unselected arrive in December with no technical interview practice for ordinary software roles and no applications in flight. Running big tech or broader technical applications in parallel costs little, because much of the preparation overlaps.
Training mental arithmetic
The most directly assessed and most reliably neglected skill. Two months of daily practice moves nearly everyone substantially, and the drills are specific.
Two-digit multiplication. 37 × 48, quickly, without writing. Learn the decomposition tricks — 37 × 48 as 37 × 50 minus 37 × 2 — until they are automatic rather than deliberate.
Fractions to decimals. Know the common ones cold: sevenths, ninths, elevenths, sixteenths. Interviewers use them constantly because they separate people who have practised from people who have not.
Percentages both ways. 17% of 240, and 41 as a percentage of 65. Estimation is acceptable and expected; state that you are estimating.
Powers and roots. Squares to about thirty, cubes to about fifteen, and rough square roots by interpolation.
Compounding. 1.05 to the fourth power, approximately, without a calculator.
Expected value on the fly. Given a payoff structure, compute the fair price in seconds.
How to practise. Fifteen minutes a day, timed, with a mix. Use a drill app or generate problems yourself. The point is speed under a clock, not accuracy at leisure — practising untimed builds the wrong skill, exactly as it does for psychometric tests.
Track your times. You should see a measurable improvement over eight weeks, and seeing it is what sustains the habit.
Probability, at the level actually asked
The mathematics is undergraduate and the fluency required is higher than most courses build. These are the patterns that recur.
Expected value of a game. Given payoffs and probabilities, compute the fair price. Then the follow-up: would you pay that price, and how much would you size the bet?
Conditional probability. The classic problems — two children, three doors, a positive test result for a rare condition. The point is not the trick; it is whether you can set up the conditioning correctly under pressure.
Sampling without replacement. Cards and urns, and whether you can reason about dependence cleanly.
Optimal stopping. You draw numbers sequentially and must decide when to stop. Common, and the reasoning is more important than the closed form.
Variance and risk. Not just the expected value but the spread, and whether you would take a bet with positive expectation and ruinous downside. The right answer is frequently no, and saying so demonstrates exactly the judgement being tested.
Symmetry arguments. Many problems that look computational collapse to a line of reasoning about symmetry. Spotting that is what separates fast candidates from slow ones.
How to prepare. Work through a standard probability puzzle collection, but with a specific discipline: solve aloud, state your setup before computing, and time yourself. Being able to reach the answer eventually is not the skill; reaching it while explaining is.
And know when to say the honest thing. "My instinct is that this is around a third, and let me check that properly" is a good answer. Silence followed by a confident wrong number is not.
What the work is actually like
Worth knowing before you spend a year preparing, because the job is narrower than the compensation suggests and it suits a specific temperament.
Trading. Market hours dominate your day, and they are early. Intense during the session, quieter after the close. You are measured constantly and precisely, which some people find clarifying and others find corrosive. You will be wrong frequently and publicly, and the ability to be wrong without it damaging your next decision is the actual job requirement.
Research. Closer to academic work with a commercial deadline. Long stretches of analysis, most ideas fail, and the feedback is slower than in trading but still unambiguous.
Development. Software engineering with unusual constraints — correctness matters absolutely, latency matters enormously, and the systems are complex. The closest of the three to an ordinary engineering job, with more pressure.
Across all three. Small teams, flat structures, and very little tolerance for imprecision in conversation. Hours are long but generally more contained than in investment banking. Compensation is high and variable, and job security is lower than the salary implies — performance is measurable, so underperformance is visible.
Who thrives. People who genuinely enjoy problems with definite answers, who are competitive without needing to be right, and who can separate a bad outcome from a bad decision.
Who does not. People who need to see the human consequence of their work, who want breadth, or who find continuous measurement stressful rather than motivating.
The firms, and how they differ
Candidates treat the sector as one employer. It is not, and the differences change where you should apply.
Proprietary trading firms. Trade their own capital. Flat, small, engineering-heavy, and the highest-paying at entry. Recruit hardest on raw problem-solving and speed. Little client contact and no sales function.
Market makers. Provide liquidity across venues, frequently at enormous volume. Similar assessment profile, more emphasis on systems and risk management at scale.
Quantitative hedge funds. Systematic strategies driven by research. Weight research ability and statistical depth more heavily, and frequently prefer PhDs for research seats while hiring engineers broadly.
Bank trading desks. More structured, more regulated, more client-facing, and generally lower paid than the specialist firms — with better-defined training programmes and a more conventional career ladder. The process resembles a superday more than a puzzle marathon.
Smaller proprietary shops. Less known, hire a handful of people, and receive a fraction of the applications. Frequently excellent places to learn, and consistently overlooked because students only apply to the names they have heard on forums.
The practical instruction: build a list of fifteen to twenty firms rather than the four everyone names. The assessment style is similar enough that preparation transfers completely, so each additional application costs almost nothing and the smaller firms have far better ratios.
Common Mistakes
- Applying in the final autumn. The main pipeline runs a year earlier through internship conversion.
- Preparing for the wrong role. Trader, researcher and developer assessments differ substantially.
- Neglecting mental arithmetic. The most trainable assessed skill and the most commonly skipped.
- Solving silently. Reasoning aloud is scored, and silence scores nothing.
- Treating rejection as a verdict. Intakes are tiny and the base rate is overwhelming.
- Having no parallel track. Preparation overlaps substantially with ordinary technical interviews; running both costs little.
Market-making games, explained
The most distinctive part of trader interviews and the one candidates have never encountered before.
The format. The interviewer names something uncertain — the number of a certain object in the room, a statistic, the outcome of a dice game — and asks you to make a market: quote a price you would buy at and a price you would sell at. They then trade against you and reveal information as you go.
What is assessed. Whether your midpoint is a sensible estimate. Whether your spread reflects your uncertainty — wide when you know little, tighter when you know more. Whether you update correctly when the interviewer trades against you. And whether you stay composed when you discover you were wrong.
The key insight most candidates miss. If someone buys at your offer, that is information: they think it is worth more than you quoted. A trader adjusts upward. A candidate who leaves their market unchanged after being hit repeatedly is signalling that they do not understand what just happened.
A worked shape. Asked to price the total number of legs in the building. You reason aloud: roughly forty people, mostly two legs, plus chairs at four each, maybe sixty chairs. Estimate around 320. You quote 280 at 360 — a wide spread, because your estimate is rough. The interviewer buys at 360. You revise: perhaps you undercounted chairs. You requote higher.
How to practise. With a partner, twenty minutes at a time. The first three attempts are always poor and the improvement after five is dramatic, which is exactly why doing them before the interview matters.
And expect to be pushed. Interviewers deliberately trade against you when you are right, to see whether you fold. Holding a position you can justify — while genuinely updating when the argument is good — is the behaviour being tested.
Getting in from outside the usual schools
Firms recruit heavily at a small set of institutions with strong mathematical programmes, and there is a genuine open route that does not depend on being at one.
Open online assessments. Several firms run tests that anyone can take, and performance is the filter. This is the most meritocratic entry point in graduate hiring anywhere — the test does not know where you study.
Competitions. Trading competitions, mathematical olympiads, programming contests and prediction tournaments all generate credible evidence, and firms watch some of them directly.
Puzzle and research sites. Some firms publish problems and hire from the people who solve them well.
Cold applications with evidence. A short message with a concrete demonstration — a competition placement, a piece of quantitative work you can show — reaches further here than in most sectors, because the assessment is objective and the firms are genuinely hunting for ability wherever it sits.
What matters more than the institution: demonstrable performance under timed conditions. That is unusual, and it means the non-target playbook applies with an extra advantage — the assessment itself is your channel, and no referral is required to take a test.
Apply widely across the sector. There are more firms than the three or four everyone names, including smaller proprietary shops that hire well and attract a fraction of the applications.
A twelve-week preparation plan
Enough to be genuinely competitive if you start early. Roughly an hour a day.
Weeks 1–3: arithmetic foundations. Fifteen minutes daily on timed mental arithmetic, plus forty minutes on undergraduate probability from a standard text. No puzzles yet — build the fluency the puzzles assume.
Weeks 4–6: probability patterns. Work through a puzzle collection, solving aloud with a stated setup before computing. Keep the daily arithmetic running; it is the base skill and it decays.
Weeks 7–8: timed conditions. Full mock online assessments under a real clock. This is where most candidates discover that they can do the mathematics and not at the required speed, which is a training problem rather than an ability problem.
Weeks 9–10: market making and games. With a partner, twenty minutes at a time. Take turns being the interviewer. This is unfamiliar and it improves faster than anything else on the list.
Week 11: coding, weighted to whether you are targeting developer or research roles. The standard technical preparation applies at a higher bar.
Week 12: consolidate and rehearse. Mock interviews end to end, and practise being pushed back on when you are right.
Throughout: know your own resume technically, because anything on it will be probed to depth.
And start this in your penultimate spring, not your final autumn. The internship pipeline is where the graduate seats come from, and by the time most students begin, that round has already closed.
Handling the interview itself
The behavioural conventions here differ from ordinary graduate interviews in ways worth knowing.
Think aloud, always. Silence scores nothing. Even "I'm not sure yet, let me consider the two cases" is better than a pause.
State assumptions explicitly. "I'll assume the die is fair and the draws are independent" — interviewers frequently intend an ambiguity and want to see whether you notice.
Estimate before computing. "This should be somewhere around a third" then verify. It demonstrates number sense and protects you if the arithmetic slips.
Say the honest thing when you are stuck. "I know this is a conditional probability and I'm setting the conditioning up wrong" invites a hint and shows you understand the shape of the problem. Interviewers here are generally happy to help a candidate who is reasoning.
Do not defend a wrong answer. Update fast, say what changed your mind. The willingness to abandon a position on evidence is a core competency in this industry and it is tested deliberately.
Do not fold when you are right either. Interviewers push back on correct answers. "I've thought about your point and I still think it's a third, because..." is exactly the behaviour they want.
Be precise in language. This industry is unusually intolerant of vague phrasing. Say what you mean, and if you are approximating, say that you are.
And treat being wrong as normal. Everyone is wrong frequently in this work. What is assessed is whether being wrong disrupts your next decision.
Frequently Asked Questions
Do I need a maths or physics degree?
For trader and researcher roles, a strong quantitative background helps considerably, and computer science, engineering and economics candidates are hired too. For developer roles, computer science is the norm and the mathematical bar is lower.
Is a PhD required?
For many quantitative researcher positions, effectively yes or close to it. For trading and development roles, no — these hire at bachelor's and master's level.
How much does the school matter?
More than in general software hiring and less than in banking. Firms recruit heavily at a set of institutions with strong mathematical programmes, and they also run open online assessments that anyone can take — which is the route in from outside, and worth combining with the non-target approach.
Is the pay really that high?
Entry compensation at the top firms is among the highest available to graduates, with a large variable component. It comes with narrow work, intense assessment culture, and less job security than the number suggests.
Can international students get these roles?
Larger firms sponsor routinely; smaller ones vary. Ask early, and note that the same enrolment and eligibility questions apply as anywhere else.
What if I fail the online assessment?
Most firms allow reapplication after a defined period, commonly six to twelve months. Use the interval to train arithmetic and probability specifically, since that is almost always what the test measured.
What is your resume scoring right now?
Scan it against a job description and get your ATS match score in about a minute.
Drop your resume here or choose a file
PDF only. Max 2 MB.
We never share your data or use it to train AI models.
See what your portfolio could look like
This theme is built straight from your resume — pick one and your portfolio is live in minutes.
Was this guide useful?
Be the first to rate it.
