Citadel Superday Prep

Probability and statistics round

Typical scenarios. Multi-stage random walks with absorbing boundaries, order statistics, asymmetric coin-pattern games

Common failure modes. Jumping to an intuitive answer without the sample space; not noticing a probability exceeding 1

Tactical advice. Define random variables and the sample space before writing equations; name the property you want to exploit.

Brainteaser / logic round

Typical scenarios. Combinatorial optimization (25 horses variants), asymmetric-information betting, discrete grid constraints

Common failure modes. Silent solving; refusing to pivot when told an approach is a dead end

Tactical advice. Solve a simplified base case (N=3 instead of 100) and look for an inductive pattern, voicing each step.

Mental-math / arithmetic-speed round

Typical scenarios. Rapid fraction division, two-digit squares, roots to two decimals, EV with fractions

Common failure modes. Freezing after one error; dropping trailing zeros; failing to round when asked to estimate

Tactical advice. Drill for weeks; use algebraic expansion such as 43 times 47 = 45^2 - 2^2 = 2021.

Live coding / algorithms round

Typical scenarios. A custom O(1) rolling-VWAP/volume aggregator, order matching, DP or graph-coloring optimization

Common failure modes. No edge-case checks (nulls, overflow, empty inputs); failing to state or optimize Big-O

Tactical advice. Explain and get sign-off on your strategy before writing a line; use clean names and defensive loops.

System design round (SWE / senior QR-QT)

Typical scenarios. A market-data normalizer ingesting 50 exchange feeds; a risk gateway clearing orders in under 5 microseconds

Common failure modes. A generic web-scale answer (message broker, slow database) where microsecond hardware is required

Tactical advice. Go straight to hardware bottlenecks, UDP vs TCP, lock-free queues, memory-mapped files and cache locality.

Live trading / market-making game

Typical scenarios. Quote a hidden or upcoming metric (dice sum, windows in the building across the street) as the interviewer trades on asymmetric information

Common failure modes. Over-exposing inventory and getting cleaned out; spreads too tight when uncertainty is high; not shifting mid after being hit

Tactical advice. Treat it as inventory optimization: if you buy, lower both bid and offer to flatten; widen spreads when uncertainty spikes.

Research or markets case

Typical scenarios. Defending an ML model and feature engineering, or a full long/short equity thesis with valuation

Common failure modes. Defensiveness when a flaw is found; fabricating data points

Tactical advice. Acknowledge data limits; if a methodology flaw is raised, propose a rolling out-of-time validation window.

Behavioral / PM / desk-head round

Typical scenarios. Past failures, research-team conflicts, why Citadel over competitors, thriving under stress

Common failure modes. Engineered generic answers, arrogance, no understanding of what Citadel does

Tactical advice. Be direct and concise; explain a mistake clearly, state the lesson and how you applied it commercially.

Informal lunch / social with the team

Common failure modes. Letting your guard down completely, unprofessional remarks, entitlement toward staff

Tactical advice. Ask about daily workflows and how ideas move from research to production; peers file qualitative feedback afterward.

Quantitative Researcher (virtual superday)

Prep. Four 60-minute technical rounds on Zoom and CoderPad.

Experience. The first two were pure math and stats, including expected steps for a 3D random walk returning to the origin; I stumbled on the joint probabilities but took a symmetry hint and vocalized the adjustment. Live coding required a rolling rank filter, which I implemented with two balanced heaps for log-time updates. The final PM round picked apart my thesis feature selection line by line.

Outcome. I never got defensive, detailed the validation data and acknowledged model limits; a competitive full-time offer came two days later.

Quantitative Trader (in-person, Miami HQ)

Prep. Rapid mental-math and probability check, then a live trading game with two senior execution traders.

Experience. I got caught long early as the interviewer aggressively hit my bid and started losing simulated money; I realized I was not dropping prices fast enough, widened spreads and skewed down to flatten and recover P&L. After an informal lunch with two junior traders, a blunt behavioral round with the Head of Credit Trading probed a prior internship failure, which I laid out without excuses.

Outcome. My recruiter called the next morning at the airport to extend the offer.