Citadel Securities sits at the deeper end of the quantitative bar among market-making firms. Candidates consistently describe interviews where a clean derivation beats a clever guess, where "what is the distribution" is a real question rather than a rhetorical one, and where the follow-up to a correct answer is usually a harder version of it. If Optiver's first filter tests arithmetic speed and SIG's rounds test decision quality, the Citadel Securities loop tests whether the mathematics is genuinely yours.
Standard disclaimers, stated plainly. QuantPit is independent and unaffiliated with Citadel Securities. What follows reflects patterns candidates report publicly and repeatedly, not an official syllabus, and nothing here reproduces a real question. Processes differ by role (trading, quantitative research, software), office, and cycle. Our Citadel Securities firm page pairs this with practice sets in the same style.
The shape of the process
Candidates describe a funnel that varies most at the front and converges at the back.
The first filter is usually an online assessment or a technical phone screen depending on role and region. Trading candidates report probability-weighted assessments and quantitative screens; research candidates report harder mathematics and sometimes coding; software candidates get standard algorithmic screens. The formats are catalogued in our assessment formats guide.
The middle rounds are technical conversations with practitioners: probability, statistics, and market intuition, with the difficulty ratcheting on follow-ups. Expect to be asked why your method works, not just whether you can execute it.
The final stage is typically several interviews close together, mixing the technical core with market questions, occasional estimation, and behavioral discussion. Reports frequently mention questions that begin simply and are then perturbed: change an assumption, remove a symmetry, extend to n steps.
The four technical areas that carry the loop
1. Probability with process flavor
This is the area candidates most associate with the firm. Beyond static probability, expect questions about things that evolve: random walks, hitting times, gambler's ruin, symmetric-random-walk arguments, and problems where the elegant solution is a conservation or martingale-style argument rather than brute force.
The skill worth building is recognizing when a quantity does not drift. If a random process has the property that its expected next value equals its current value, then averaging over the whole path collapses to a single equation, and problems that look like they need heavy machinery fall out in three lines. Practice the underlying reasoning in /topics/probability/, then push into harder tiers where the multi-step and recursive problems live.
2. Statistics that you can actually derive
Candidates report genuine statistics rather than trivia: variance and covariance manipulations, estimator properties (bias, consistency, efficiency), regression intuition including what the coefficients mean when predictors are correlated, and the standard traps around correlation, causation, and selection. The statistics topic set covers the ground; the bar is being able to derive rather than recite.
3. Options and market structure
For trading roles, reports describe put-call parity, the intuition behind the Greeks, why delta hedging works and what it costs, and how volatility enters a price. You do not typically need to reproduce a full Black-Scholes derivation, but you do need to reason about what happens to a position when a parameter moves. Work through /topics/options/ until the sensitivities are intuitive rather than memorized.
4. Markets fluency
The distinguishing question in many reports is some version of "how does this connect to real trading?" Candidates who can talk about liquidity, spreads, adverse selection, and why market makers earn what they earn separate themselves from candidates who can only compute. Our market making topic set builds the vocabulary, and the market maker game makes the adverse-selection lesson concrete by charging you for it.
How the questions escalate
The most commonly reported pattern is the ladder: a clean question, then a perturbation, then a generalization. Two coins becomes n coins. A symmetric walk becomes a biased one. A fair die becomes a die you may re-roll once, then twice, then optimally.
That structure tells you what to practice. Solving one instance is not enough; you want to solve it in a form that survives generalization, which usually means setting up a recursion or an invariant instead of enumerating cases. When you finish a practice question, the highest-value next step is to ask what the interviewer would perturb, then solve that version too.
A preparation plan
- Weeks 1 to 2: rebuild the core. Conditional probability, expectation and variance rules, combinatorics. Easy and medium tiers, until the algebra is automatic. See the full sequence in the prep roadmap.
- Weeks 3 to 4: process problems. Random walks, recursive expectation, hitting times, ruin problems. These are the differentiator, and they reward practice more than talent.
- Week 5: statistics and options. Estimators and regression intuition; parity and the Greeks by reasoning rather than memory.
- Throughout: markets vocabulary. Ten minutes a day of reading about market structure compounds into the fluency the final rounds probe.
- Final week: mocks. Timed, spoken, and with a partner or the mock interview mode that grades under a clock.
FAQ
How hard is the Citadel Securities interview compared to other trading firms?
Candidates generally describe a higher mathematical bar than the arithmetic-screen-first firms, with more emphasis on derivations, stochastic-flavored probability, and statistics. Speed still matters, but depth is the differentiator most reports emphasize.
Do I need stochastic calculus?
For most trading roles, no: the reported questions live in discrete probability, random walks, and expectation arguments rather than continuous-time machinery. Quantitative research roles can go further. Understanding martingale-style reasoning intuitively is more useful than formal measure theory for the trading track.
Are there brainteasers?
Yes, but reports suggest they function as openers rather than the main event, and the follow-ups pull them toward real mathematics. Treat a brainteaser as the first rung of a ladder, and expect to be asked to generalize it.
Is there a coding component?
For quantitative research and software roles, commonly yes. For trading roles it varies by region and cycle; check the posting. Our assessment formats guide covers what the coding screens tend to look like.
What separates candidates who advance?
Three things in the reports: deriving rather than recalling, handling the perturbation follow-up without losing composure, and connecting the mathematics to how markets actually work. The third is the one most candidates neglect, and it is trainable in a couple of weeks.