Optiver Quantitative Research Interview Guide

Optiver's Amsterdam graduate QR posting includes online assessments. Prepare for quantitative reasoning, coding and research discussion, while checking the stages for your particular vacancy.

Content and official sources reviewed · 25 September 2026 · Official sources and scope

What should you expect from Optiver quantitative research recruiting?

Start with the vacancy, not a memorised list of rounds. Optiver's current Amsterdam graduate research posting includes an online assessment, while its research roles place different demands on academic background and experience. The official pages reviewed here do not establish one worldwide quantitative research interview sequence. Prepare to solve quantitative problems, write clear code and defend research decisions, then adapt that preparation to your invitation.

How to read this guide: official requirements are linked and scoped to their vacancy or region. Round order, assessment content and any take-home work are treated as variable. Exercises and preparation advice are EverythingQuant recommendations, not official or recalled Optiver questions. Your recruiter and current invitation take precedence over an older guide or candidate account.

Process at a glance

Use this as a preparation map, not a five-round promise. Ask which stages apply before deciding how much time to spend on assessment speed, programming or a project presentation.

Eligibility: distinguish the research role you are applying for

A PhD is not a universal entry requirement

The Amsterdam graduate QR vacancy for a 2027 start asks for a relevant quantitative degree, programming and mathematical problem solving; it does not state a universal PhD requirement. In contrast, the New York graduate QR PhD vacancy explicitly targets PhD candidates. The Amsterdam 2027 research internship specifies penultimate-year students graduating in 2028.

Before applying, record the exact job title, office, degree requirement, graduation window, start date and any work-eligibility conditions. A similarly named role in another country is not evidence that you qualify for this one. Experienced and specialised research roles can set different expectations again.

Application: make the evidence of research ability easy to find

EverythingQuant recommendation: give each important project a compact evidence trail: the question, the data, your contribution, how you evaluated the result and one limitation. “Built a machine-learning model” says less than “compared a regularised model with a simple baseline using a chronological holdout, then investigated why performance weakened out of sample.” Only include results you can explain and reproduce.

Separate your own work from a team's work. For a thesis, describe the methodological contribution rather than assuming the title explains it. For an industry project, respect confidentiality: explain the design and evaluation without disclosing restricted data or proprietary details. Select one project you can discuss deeply and a second that demonstrates a different strength.

Online assessment: follow the role-specific instructions

Official, vacancy-specific: the reviewed Amsterdam graduate QR posting asks selected applicants to complete its online assessment within seven days. That is not a universal deadline or a published specification of every Optiver research test. Check the invitation for the actual deadline, permitted tools, language, timing, accessibility arrangements and practice instructions.

Do not copy the trading assessment battery onto QR

Names such as 80 in 8, NumberLogic, Beat the Odds and Zap do not, by themselves, establish what your research application contains. This guide no longer presents that collection as mandatory QR stages. When the invitation names a format, practise that format; when it does not, prioritise mathematical reasoning and coding rather than preparing every trading tool.

A useful diagnostic is to solve a few probability problems untimed, then repeat a comparable set with a clock. Separate conceptual mistakes from slow arithmetic or misread instructions. An answer you can produce quickly but cannot justify is weak preparation for a live research discussion.

Coding and data analysis: make correctness visible

Optiver's Amsterdam graduate posting names programming, with Python preferred; its research overview describes work from hypotheses and data selection through testing and implementation. Those descriptions support preparing both code and research judgement. They do not establish that every candidate receives the same coding platform or a take-home dataset.

A useful practice task

EverythingQuant exercise: given timestamped observations, compute a rolling statistic and evaluate a simple predictor on later observations. Before coding, specify how you handle missing values, duplicate timestamps, insufficient history and an empty input. Show that an observation at time t cannot use data that becomes available after t. Then explain the time and memory costs of your implementation.

A simple, tested solution is a better starting point than a complicated model with unclear input assumptions. Practise explaining your approach while coding, testing boundary cases and correcting a mistake without losing the thread of the discussion.

For systematic study, Explore the data structures and algorithms course . Its overview is public; enrolment and lesson permissions remain in the course. This develops coding foundations, not a complete Optiver research assessment.

Recruiter discussion: establish the format and your motivation

Explain why research, not just why quantitative finance

Prepare a specific account of what you enjoy investigating: identifying an uncertain relationship, designing a test, understanding a failed result or improving an existing model. Connect that preference to the work described in the vacancy. You do not need to pretend that all of your previous projects were trading projects.

Clarify the next stage

Ask whether you should expect live coding, a mathematical interview, a research presentation or another exercise. If presenting, confirm the audience, time limit and whether slides or a repository are wanted. The Europe recruitment FAQ is useful regional guidance, but your recruiter should resolve vacancy-specific questions.

Technical research interviews: reason beyond the final number

EverythingQuant preparation priorities: conditional probability, expectation and variance, estimation, uncertainty, basic linear algebra and model evaluation. Practise identifying what is known, what is assumed and what additional observation would change your answer. These are preparation topics, not a guaranteed interview syllabus.

Need to rebuild the concepts first? View the course introduction . The public introduction explains the course; later lessons retain their enrolment and paid-access requirements.

How to discuss a research project

Explain the question, baseline and assumptions

Begin with the decision or quantity you wanted to predict. Explain why the available data could answer the question and what your baseline was. State the assumptions most likely to fail. Describe why you chose the method over a simpler alternative, not just why the method is fashionable.

Show how you ruled out leakage and overfitting

For a time-dependent prediction problem, describe a split that respects when information became available. Fit transformations using training data only. Keep tuning separate from final evaluation. If adjacent labels overlap in time, explain how your evaluation avoids giving the training set information about the test period; the appropriate separation depends on the prediction horizon and task.

Explain how many variants you tried and how you selected the final one. A single attractive holdout result after repeated inspection is not an untouched test. For a non-temporal problem, justify the relevant grouping or independence assumptions instead of applying a time-series recipe automatically.

Connect the metric to the objective

Explain why your metric measures the intended task, how it compares with the baseline and where performance is unstable. For a model intended to support trading, distinguish predictive performance from implementable returns after costs and constraints. For another research task, use its actual objective rather than forcing every result into a trading metric.

Present decisions and limitations, not a wall of results

Practise a two-minute summary, then a deeper version covering data, method, validation, failure analysis and next steps. Be ready to explain one decision you changed because of evidence. “This result is promising, but the sample is small and the effect is unstable in this regime” is more useful than defending an unjustified claim of success.

Later and final rounds: prepare for the confirmed agenda

Do not assume a universal “superday,” mandatory take-home project or in-house trading simulation from this guide. The reviewed sources do not establish those as compulsory stages across QR roles. If an exercise is assigned, ask about the time expectation, deliverables, permitted tools and how it will be discussed. If no exercise is assigned, do not invent one at the expense of the actual interview.

Prepare to revisit an earlier answer, explain a changed assumption and discuss your work with people from different technical backgrounds. A practice partner can challenge your baseline, ask how a result could be wrong and request a simpler explanation. The aim is to make your reasoning inspectable, not to memorise a performance.

How to allocate your preparation

Before applying: audit eligibility and choose the projects that best show independent thought. Assessment received: diagnose the confirmed topics and practise the stated timing. Technical interview scheduled: mix probability reasoning with live coding and a project defence. Presentation requested: rehearse the exact format and reserve time for questions about limitations.

Keep an error log with three columns: the mistake, why it happened and a different problem that would test the same weakness. Avoid completing large numbers of trading puzzles merely because they carry the Optiver name.

This action carries Optiver + Quantitative Research as guide context. Confirm those choices in the existing questionnaire: automatic prefill is not assumed, and following the link does not replace an existing preparation profile or roadmap.

Preparation checklist

• I have checked the actual vacancy and invitation, including the office, degree requirement and deadline.
• I can solve a probability problem, state its assumptions and interpret the answer.
• I can write tested code and explain boundary cases, complexity and information timing.
• I can defend a project's baseline, validation, metric and limitations without overstating the result.
• I know which next stage I am preparing for—and which supposed stages remain unconfirmed.

Sources reviewed on 25 September 2026. Vacancy examples describe their named locations and starts, not guaranteed availability or a worldwide process. EverythingQuant is independent of Optiver.