GSA Capital Interview Process & Prep
GSA Capital is a London-based systematic hedge fund rooted in Deutsche Bank's Global Statistical Arbitrage group. Strong intel: timed OA replica, worked interview reports, past take-home (data, code).
Quant Researcher applicants first clear a 60-minute, 25-question Correlation One online assessment (ML/statistics, numeric probability and linear algebra, brainteasers, behavioral finance). Then comes the signature take-home data project: predict instrument Y's 10-minutes-ahead returns from X/Y prices, judged purely on out-of-sample MSE. Technical interviews stress regression theory, statistics, portfolio construction and (recently) machine learning.
Practice every firm's OA: timed replicas, free game replicas and guides in one place →
The GSA Capital interview funnel
1. Online Assessment — Correlation One (Quant Researcher)
Correlation One · 25 questions / 60 min · MCQ + numeric entryQuant Researcher first screen, Correlation One: 25 questions in 60 minutes, single-option MCQ or numeric entry, PREVIOUS / NEXT navigation, 25-box progress bar. OA page: timed replica, verified answers.
Aug 2026 sitting, four families. ML/statistics: k-means local optima, overfitting, residual-vs-one-variable plots, R-squared vs t-statistics, regression on 110 observations, 100 variables. Probability and linear algebra ('integer closest to 20p', 'M + N for M/N in lowest terms'): P(3X + 4Y > 5), eigenvalue sum, null-space dimension, random quadratic roots, two sample means, 8-coin Bayes, two cards.
Brainteasers: nine dots in four segments, molecule-collision invariant, knights and knaves. Behavioral-finance judgment: cab-witness base rate, Linda the bank teller, Wason four-card task, 2-4-8-16 probe, rural-hospital extremes, three 'which statement do you agree with more' pairs (sunk cost, process vs outcome, survivorship).
2. Take-Home Data Project
Take-home · own environment · graded on OOS MSEGSA's signature stage. From real 10-second-bar prices for instruments X and Y, predict Y's 10-minute-ahead return with data up to t only, in two template functions (modelEstimate fits parameters; modelForecast, one prediction per row). Graded purely on out-of-sample MSE: clean and validated beats complex; lookahead / leakage is fatal.
3. Technical — Regression & Statistics
45–60 min · with a researcherThe core technical round: regression theory and statistics. Expect ridge closed form and why X'X + λI is invertible, variance of a sum of equicorrelated variables (and valid-ρ bounds), the bus-stop inspection paradox, years-until-positive-return via the CLT, Sharpe-ratio backtest length, and estimator reasoning. They want a clean argument, not just a number.
4. Portfolio Construction & Machine Learning
45–60 min · with a researcherA round on portfolio optimization and, increasingly, machine learning. Expect Markowitz optimization given an alpha signal and covariance matrix (objective, constraints, unconstrained closed form), covariance-estimation methods (sample vs EWMA vs PCA factor models vs graphical Lasso), and recent ML items: PCA components with near-zero R², tree feature importance / SHAP values, and neural-net initialization / batch-norm-and-dropout at test time.
GSA Capital Portfolio Construction & Machine Learning interview guide →
5. Final — Research Fit
Conversational · with senior researchersA closing research-fit stage: a deep-dive on your take-home methodology and past projects, plus culture signal. They probe why you made each modeling choice, how you validated it, and how you'd extend it.
GSA Capital interview — FAQ
What is the GSA Capital interview process?
GSA Capital is a London-based systematic hedge fund rooted in Deutsche Bank's Global Statistical Arbitrage group. Strong intel: timed OA replica, worked interview reports, past take-home (data, code). Quant Researcher applicants first clear a 60-minute, 25-question Correlation One online assessment (ML/statistics, numeric probability and linear algebra, brainteasers, behavioral finance). Then comes the signature take-home data project: predict instrument Y's 10-minutes-ahead returns from X/Y prices, judged purely on out-of-sample MSE. Technical interviews stress regression theory, statistics, portfolio construction and (recently) machine learning. The loop runs 5 stages: Online Assessment — Correlation One (Quant Researcher), Take-Home Data Project, Technical — Regression & Statistics, Portfolio Construction & Machine Learning, Final — Research Fit.
How many rounds does GSA Capital have?
5 stages in total, starting with the Online Assessment — Correlation One (Quant Researcher) and ending with the Final — Research Fit.
How do I prepare for the GSA Capital interview?
Work the stage notes above, then drill the GSA Capital interview-questions set and the GSA Capital online-assessment practice — each problem has a full worked solution.
GSA Capital round-by-round guides
GSA Capital interview guide: the firm, culture, roles and pay →
Stage guides for every firm: superdays · behavioral questions · market-making games