Back to Citadel guide

Citadel · Online Assessment

Citadel Online Assessment Prep

Citadel screens candidates through HackerRank and CodeSignal (plus custom/Tradermath-style trading sims) before any interview. Below: the exact format, what each section tests, and how to practise it in identical conditions until you walk in confident.

Practise for Citadel

Freeno card

Start practising on Intervyo. Free tools, scored feedback, no payment.

  • Resume Checker, scored against Citadel
  • HireVue practice, AI-scored
  • Live AI mock interviews with Vyo
  • Psychometric tests in real formats
  • Application Tracker
Start practising free
Every interview stage, with AI feedback
Upgrade any time, no commitment

The format

What Citadel's online assessment actually looks like

The first highly mechanized filter, often auto-triggered within 48-72 hours of applying, frequently before any human resume review.

Timed sections

Most online assessments split into 3-5 sections, each with its own clock. Speed and accuracy both count.

Adaptive difficulty

Modern formats get harder if you answer correctly, easier if you struggle. Your final score reflects what you can actually do under time pressure.

Pass mark

Citadel sets a pass mark per test type. Below it, you don't progress regardless of how strong the rest of your application is.

Completion window. Typically 3-5 days to launch and complete once the link arrives; component timers inside are rigid.

By division. SWE is coding-only; QR adds advanced probability and statistics; QT is the most multi-faceted (speed math, sequences, probability puzzles, market-making sim); Investing/Fundamental leans on data and financial analysis plus a behavioral questionnaire.

Recent changes. Adaptive question banks, advanced proctoring (copy-paste and window-switch flags), and explicit market-making mini-games embedded in the QT battery.

The provider

What Citadel actually buys

Citadel configures its own selection of HackerRank and CodeSignal (plus custom/Tradermath-style trading sims) modules. Below: the exact products in the suite, why they were chosen, and what the provider's reputation is across the candidate pool.

Modules in the suite

  • CodeSignal General Coding Assessment (GCA)
  • HackerRank coding modules with strict complexity limits
  • Custom browser-based mental-math and number-sequence drills
  • Interactive market-making / trading simulation

History at Citadel. An automated gatekeeper that filters hundreds of thousands of applicants on speed and accuracy before any phone screen.

Candidate reputation. Demanding mainly because of extreme time pressure and negative marking on the speed modules; components do not compensate for one another.

Section breakdown

What each part of the Citadel assessment tests

Each section has its own format, timer and trap pattern. Worked examples below show exactly what you will see and where candidates drop points.

Coding / algorithms (SWE, QR)

2-4 problems · 60-120 minutes

What it tests. Correctness on all hidden cases, execution speed and optimal time/space complexity

Worked example. Streaming maximum in a sliding window of size K, solved in O(N) with a monotonic deque rather than the naive O(N*K).

Common traps. Integer overflow from 32-bit types, off-by-one boundaries, and TLE from suboptimal complexity.

How to handle it. Check input limits first to deduce the target complexity; use 64-bit integers; pre-size outputs.

Mental math speed (QT)

Single-digit minutes (for example 80 questions in 8 minutes)

What it tests. Immediate, accurate arithmetic processing

Worked example. 43 times 37 via difference of squares: (40+3)(40-3) = 1600 - 9 = 1591.

Common traps. Negative marking punishes guessing; a single error can break rhythm.

How to handle it. If you cannot solve within 4-5 seconds, skip cleanly; aim for high accuracy across high volume.

Number sequences / patterns

What it tests. Rapid inductive reasoning on sparse data

Worked example. For 12, 14, 18, 26, 38, 62 the rule is add the product of the digits, so 62 + (6*2) = 74.

Common traps. Patterns scale beyond linear/quadratic - interleaved sub-sequences and digit manipulations.

How to handle it. Test differences and products of digits; deconstruct odd/even index sub-sequences.

Probability, statistics, combinatorics (QR, QT)

What it tests. Conditional probability, discrete distributions and combinatorics under time pressure

Worked example. Optimal-play dice game: keep 4/5/6, reroll on 1/2/3, so EV = 0.5(5.0) + 0.5(3.5) = 4.25.

Common traps. Spending 5-6 minutes on one combinatorics item with backward navigation disabled.

How to handle it. Time-box per question; keep clean scratch paper for state variables.

Trading / market-making simulation (QT)

What it tests. Total EV captured, inventory P&L volatility and adaptability to new information

Worked example. Trading the sum of two dice (initial EV 7.0): when the first die is revealed as 6, EV jumps to 9.5 and you must re-price immediately, skewing to manage inventory.

Common traps. Treating it as pure math and ignoring inventory; getting cleaned out by adverse selection.

How to handle it. Skew prices early and decisively to shed excess inventory, sacrificing small margin to cut risk.

Personality / behavioral questionnaire

What it tests. High risk-tolerance within structured boundaries, data-driven objectivity and resilience

How to handle it. Answer for calculated, capped risk and empirical decision-making, not an idealized empathetic profile.

Pass mark

How Citadel scores the assessment

Exceptionally selective thresholds; the OA is an aggressive filter and under 15% of test-takers advance.

Competitive percentile thresholds

  • CodeSignal GCA (SWE). Competitive raw score ~800+ / complete correctness on core problems
  • HackerRank coding (QR). 100% of visible and hidden cases with optimal big-O
  • Mental math (QT). Accuracy above ~90-95% on attempted questions at high volume
  • Trading simulation. Consistent positive EV capture with disciplined inventory risk

Methodology. Components do not compensate: a perfect coding score with a failed mental-math run is an automated rejection. Each tested attribute is a strict baseline.

Response time. Automated updates typically within 3-7 business days of completion.

Score visibility. Candidates never see raw scores, percentiles or which cases failed.

How to practise

Drill Citadel's exact format

Same provider, same section structure, same time pressure. With a scored report after every test so you can fix the weakest section before the real one.

  • HackerRank and CodeSignal (plus custom/Tradermath-style trading sims)-calibrated practice. Not a generic stand-in. The exact provider and section structure Citadel uses on the day.
  • Adaptive question difficulty. Questions get harder when you nail them, easier when you struggle. Walk away knowing where you actually stand.
  • Coaching, not just a score. "You are spending too long on table-data questions" beats "you scored 68%". Specific advice per weak section.
  • Percentile benchmarking. Your score compared to the US candidate pool, so you know if 70% is excellent or worrying for that test.

Free practice section, scored. Keep practising free on Intervyo.

Why candidates fail

How candidates lose Citadel's assessment

Specific failure patterns documented across past applicants. Most are avoidable with structured preparation.

  1. 1

    Over-engineering coding solutions

    Building elaborate abstractions for problems that need direct algorithmic logic, then running out of time on the highest-weighted questions.

  2. 2

    Timer panic on the math run

    Freezing after one error and turning a single mistake into a string of consecutive misses.

  3. 3

    Suboptimal data types

    Passing small cases but failing hidden ones via quiet 32-bit integer overflow on large inputs.

  4. 4

    Treating the trading game as pure math

    Computing exact EV while letting inventory balloon, then suffering catastrophic drawdowns on adverse outcomes.

What works

What separates the candidates who pass

Concrete habits drilled by candidates who clear the cut-off, drawn from applicant accounts and practice patterns.

  • Daily mental-math automation

    Weeks of Zetamac/Tradermath drilling so basic arithmetic frees working memory for risk decisions.

  • Strict time-boxed practice

    Rehearse under the exact CodeSignal/HackerRank time allocations to build pacing intuition.

  • Verify constraints before coding

    Read input limits (for example N up to 10^5) to deduce O(N log N) or O(N) before writing a line.

  • Proactive inventory management

    In simulations, prioritize risk mitigation: skew spreads early to shed inventory rather than chase marginal profit.

From past applicants

How recent Citadel candidates approached the assessment

Anonymised candidate accounts of how recent applicants approached the Citadel assessment. Each covers the prep, the experience, and the outcome.

Quant Trading intern applicant (Citadel Securities)

Prep. Applied out of a target undergraduate program in early September; got an OA link within 48 hours.

Experience. A high-speed arithmetic module, number series and an interactive trading game. I focused too much on exact EV early, let inventory skew to +12 contracts, and got hit by adverse selection right before a down-step. Once I started skewing - dropping my ask to encourage sales and lowering my bid to discourage buys - I controlled the position.

Outcome. Advanced to the technical phone rounds; the lesson was that inventory management matters as much as accurate EV.

Quant Research associate applicant (Citadel hedge fund)

Prep. OA arrived just three days after submitting my resume during my PhD.

Experience. HackerRank with two hard problems (a memory-tight graph traversal and a multi-dimensional DP) plus strict multiple-choice probability with no backward navigation - including a conditional-variance-after-stopping-time question and a multi-step Bayesian revision. I cleared all visible coding cases and worked the math on a notepad.

Outcome. Received a recruiter phone-screen invitation less than a week later.

Practice strategy

Where to drill the Citadel format

Structure preparation around specialized, timed resources that mirror the real component formats.

  • Targeted algorithm drills

    LeetCode Medium/Hard on sliding windows, heaps, order-book designs and graphs within tight 20-30 minute windows, plus the official CodeSignal and HackerRank practice assessments.

  • Mental-math trainers

    Zetamac, Tradermath or RankYourBrain set to double-digit multiplication, fraction-to-decimal and quick division; 15-20 minutes every morning aiming for 95%+ accuracy at rising volume.

  • Probability canon

    Work conditional probability, discrete distributions, Markov chains and games of strategy from Zhou and Crack with fast scratch work and no calculator.

  • Market-making simulators

    Practice inventory management and dynamic spread adjustment on tools like OpenQuant or EverythingQuant.

Time investment. Weeks of daily 15-20 minute speed drills plus timed full-length coding sets is typical for competitive scorers.

Time management

Five moves that protect your score

  1. 01Set your own clock per question. Divide section time by question count. Move on when you hit your per-question budget, even if you are mid-thought.
  2. 02Read the question before the chart. Half the work in numerical is finding the right number in a table. Knowing what you are looking for cuts the time in half.
  3. 03Use "Cannot Say" generously. Verbal reasoning rewards strict reading. If the passage does not say it, the answer is "Cannot Say", not your own inference.
  4. 04Skip the impossible ones. Most tests do not penalise wrong answers more than skips. If you cannot see it in 20 seconds, flag and move on.
  5. 05Practise the exact format, not a generic stand-in. HackerRank and CodeSignal (plus custom/Tradermath-style trading sims) has its own rhythm, and a generic reasoning test is not the same. Intervyo's simulation is calibrated to this format, so you rehearse the real thing under real timing rather than a generic aptitude set.

FAQ

Citadel Online Assessment questions, answered

Coding is hosted on HackerRank or CodeSignal (the GCA for SWE), with custom or Tradermath-style environments for the trading and speed modules. The environments support C++, Python, Java and more, but for Quant Research and high-frequency SWE tracks, C++ or Python align best with the core stacks - it is better to show deep, idiomatic mastery of one than a shallow grasp of several. Calculators are strictly prohibited on the math modules, though you should keep clean scratch paper for the probability and trading sections.

The other rounds

The rest of the Citadel process

Online Assessment is one of four rounds. Practise each one free on Intervyo.

HireVueLive InterviewSuperday

Practise free

Rehearse the Citadel online assessment free

Practise every stage on Intervyo with AI-scored feedback: HireVue, psychometrics, live mock interviews, and Resume. Free to start, no card required.

Start practising free

Free tools, upgrade any time

Intervyo is not affiliated with or endorsed by Citadel, SHL, Pymetrics, Cubiks, AON or any other assessment provider. Test details are sourced from past applicants and published guidance; verify on the firm's careers site before applying. Sector: Investment Banking.

Citadel

Practise free

Start free