Probability and expected value
Define the sample space and whether events are independent. Use conditional probability and Bayes’ rule when information changes. Expected value is probability-weighted payoff, but trading decisions also consider variance, tail risk, repeated play, and capital constraints.
Statistics
Questions can involve estimation, distributions, regression, confidence, hypothesis testing, bias, variance, and experimental design. State what the data can and can't identify. A significant relationship can still be economically irrelevant or unstable.
Coding and algorithms
Write readable code, clarify input and output, handle edge cases, and discuss time and space complexity. Tests should include normal, boundary, and invalid cases. Don't optimize before correctness is established.
Data questions
A data case may require cleaning, exploration, feature design, modeling, and communication. Check missingness, selection, timestamps, revisions, target leakage, and whether the sample represents the intended population.
Research design
A strong answer defines the hypothesis, data, benchmark, train and test method, transaction costs, and failure criteria. Explain why the result could be spurious and what independent evidence would increase confidence.
Market-making and trading games
These can test pricing under uncertainty, updating, inventory, and strategic behavior. State a fair value or range, adjust with new information, and understand that a quoted market creates obligations and risk.
Estimation
Break the problem into components, keep units visible, and check order of magnitude. The interviewer is evaluating model construction and communication, not hidden exact knowledge.
Behavioral questions
Prepare examples of failed research, difficult collaboration, debugging, learning a new domain, and choosing between speed and rigor. Technical teams still need judgment and communication.
Answer habits
Clarify, structure, reason aloud, test, and summarize. When an assumption changes, update the solution rather than defending the first path.
Mental arithmetic and approximation
Trading and research interviews can require ratios, percentages, logarithmic intuition, combinatorics, and quick bounds. State an approximate method and check the order of magnitude. Exact arithmetic is less valuable than a clean estimate when the question asks for reasoning.
Coding-discussion depth
After producing a working solution, be ready to discuss alternative data structures, complexity, memory, concurrency, testing, and how requirements change the design. A researcher may be asked how the code would scale to a large dataset; an engineer may be asked how the system behaves under failure.
Interpreting statistical results
A model can have a low p-value and still be unusable because the effect is small, unstable, or expensive to trade. Explain confidence intervals, sample dependence, repeated testing, and the difference between predictive and causal claims. Good answers include the result’s limitation without being prompted.