GSA Capital Online Assessment
GSA's Quantitative Researcher funnel now opens with a timed online assessment on Correlation One (assessments.correlation-one.com): 25 questions in 60 minutes, single-option multiple choice plus numeric-entry items, with PREVIOUS / NEXT navigation (you can revisit any question) and a 25-box progress bar. The Aug 2026 sitting mixed four families: machine-learning and statistics concept questions (k-means restarts, overfitting, residual plots, R-squared versus t-statistics, which regression to use when n is about p), probability and linear algebra with numeric entry (the recurring convention is 'enter the integer closest to 20p' or 'compute M + N for a reduced fraction M/N'), classic brainteasers (the nine-dots path, the molecule-collision invariant, knights and knaves), and behavioral-finance judgment items (the Kahneman cab problem, the Linda conjunction, the Wason four-card task, the 2-4-6 rule test, the small-hospitals law of small numbers, plus 'which statement do you agree with' pairs on sunk cost, process vs outcome and survivorship). After the OA comes the real take-home data project (predict the 10-minute-ahead return of instrument Y from the X/Y price history, graded on out-of-sample MSE), for which a complete past submission (data + Lasso pipeline) exists in our intel.
Practice every firm's OA: timed replicas, free game replicas and guides in one place →
Quant Researcher — Correlation One online assessment
One timed set of 25 questions in 60 minutes on Correlation One: single-option MCQs and free numeric entry ('Please enter your answer below'), PREVIOUS / NEXT navigation so you can come back to anything, and a visible countdown. Four question families in the captured sitting: (1) ML / statistics concepts, e.g. which techniques stop k-means from settling in a local optimum, what makes an overfitting linear classifier worse, which issue a one-variable-at-a-time residual plot cannot reveal, the true relationship between R-squared and coefficient t-statistics, and the right regression for 110 observations against 100 variables; (2) probability and linear algebra with a typed integer answer, using the 'integer closest to 20p' and 'M + N for M/N in lowest terms' conventions (P(3X + 4Y > 5), the sum of a 3x3 matrix's eigenvalues, the nullity of a 5x5 matrix, the chance a random quadratic has real roots, the difference of two sample means, an 8-coin Bayes update, the most likely statement about the top two cards); (3) brainteasers (nine lattice points in the fewest straight segments, the least number of molecules after collisions, a two-inhabitant knights-and-knaves puzzle); (4) behavioral-finance judgment items: the cab-witness base-rate problem, Linda the bank teller, the Wason four-card selection task, the 2-4-8-16 hypothesis-testing probe, rural-hospital mortality extremes, and three 'which statement do you agree with more' pairs (sticking with vs pivoting from a conviction, decision process vs outcome, Buffett as skill vs survivorship). Our timed replica below reproduces every captured question verbatim with a verified key; the bank problems listed here are the closest existing twins of individual items for extra practice.
- Molecule Reaction Minimization Brainteaser · hard
- Knights, Knaves, and Two Statements Logic · easy
- Posterior Probability of a Double-Headed Coin Probability · easy · free
- Eigenvalue-Determinant-Trace Relationship Linear algebra · medium
- Regression with More Features Than Observations Regression · medium
- Overfitting When Features Approach Sample Size Machine learning · medium
- Diagnosing Overfitting and Cross-Validation Machine learning · easy
- Survivorship Bias in Backtesting Finance · medium
Quant Researcher — Take-Home Data Project: 10-min return prediction
You receive data.csv with 10-second-bar prices for two instruments X and Y (timestamp in ms, xprice, yprice, and a returns column defined as the price 10 minutes ahead minus the current price, outlier-truncated). Build a model to predict Y's 10-minute-ahead return using only data up to and including time t. You must fill a fixed template: parameters = modelEstimate(trainingFilename) and predictions = modelForecast(testFilename, parameters). Extra files and external libraries are allowed; the model is judged on out-of-sample mean squared error against held-out sets of similar (not identical) shape.
- Why Use Adjusted R-Squared? Regression · easy · free
- Detecting and Addressing Multicollinearity Regression · medium
- Regression with More Features Than Observations Regression · medium
- L1 vs. L2 Regularization Regression · medium
- The Closed-Form OLS Solution and When It Exists Regression · medium
- Ridge Regression Closed Form and Why the Inverse Always Exis Regression · medium
GSA Capital OA — FAQ
What is the GSA Capital online assessment?
GSA's Quantitative Researcher funnel now opens with a timed online assessment on Correlation One (assessments.correlation-one.com): 25 questions in 60 minutes, single-option multiple choice plus numeric-entry items, with PREVIOUS / NEXT navigation (you can revisit any question) and a 25-box progress bar. The Aug 2026 sitting mixed four families: machine-learning and statistics concept questions (k-means restarts, overfitting, residual plots, R-squared versus t-statistics, which regression to use when n is about p), probability and linear algebra with numeric entry (the recurring convention is 'enter the integer closest to 20p' or 'compute M + N for a reduced fraction M/N'), classic brainteasers (the nine-dots path, the molecule-collision invariant, knights and knaves), and behavioral-finance judgment items (the Kahneman cab problem, the Linda conjunction, the Wason four-card task, the 2-4-6 rule test, the small-hospitals law of small numbers, plus 'which statement do you agree with' pairs on sunk cost, process vs outcome and survivorship). After the OA comes the real take-home data project (predict the 10-minute-ahead return of instrument Y from the X/Y price history, graded on out-of-sample MSE), for which a complete past submission (data + Lasso pipeline) exists in our intel.
What is the format of the GSA Capital OA?
It runs as 2 tracks (Correlation One online assessment, Take-Home Data Project: 10-min return prediction). Quant Researcher, Quant Researcher, each timed and auto-graded where applicable.
How do I practice for the GSA Capital online assessment?
Work the sample questions above (each with a full worked solution), then take the interactive GSA Capital OA practice on QuantVault. For the full interview, see the GSA Capital interview-questions set.