Interview Lab
Practise the reasoning, not the answers
What are quant interviews actually testing?
Quant interviews test speed and structure of reasoning under pressure, not recall. Probability and mental arithmetic appear in almost every loop, including engineering tracks. Trading loops add market-making games and estimation; engineering loops add low-level systems depth; research loops add validation discipline and how you avoid overfitting.
Firm
Topic
Stage
Difficulty
23 questions
You flip a fair coin three times. Given that at least one flip came up heads, what is the probability all three were heads?
Two people independently pick a real number uniformly from [0, 1]. What is the probability the two numbers differ by more than 0.25?
You draw cards one at a time without replacement from a standard 52-card deck until the first ace appears. What is the expected number of cards drawn?
A random walk starts at 0 and moves +1 or −1 with equal probability each step. What is the probability it reaches +3 before it reaches −5?
What is 17 × 23, without writing anything down?
Estimate the square root of 4,600 to one decimal place in under ten seconds.
A contract pays 100 if an event happens and 0 otherwise. It trades at 37. You think the true probability is 41%. You can trade 500 contracts. What is your expected edge?
I roll two fair six-sided dice and pay you the product. Make me a market, two wide.
You are quoting a market and get lifted five times in a row on the offer, in full size, by the same counterparty. Nothing else in the market has moved. What do you do?
Make a market on the number of petrol stations in the United Kingdom.
You backtest 200 signals and the best one has a Sharpe ratio of 2.4. Should you trade it?
Your alpha shows a strong in-sample relationship with next-day returns, but the effect halves out of sample. Give three explanations that are not overfitting.
Two return series each have a correlation of 0.6 with a third. What is the tightest bound you can put on their correlation with each other?
Design a data structure that supports insert, delete and get-random-element, all in expected O(1).
You receive a stream of order-book updates and must report the maximum trade price in the last N events at every step. How?
How would you implement a limit order book supporting add, cancel and match, and what is the complexity of each?
Why might std::unordered_map be a poor choice on a latency-critical hot path?
Your median tick-to-trade latency is fine but the 99.9th percentile is ten times worse. Name the likely causes.
Explain false sharing and how you would detect it in a multi-threaded market data handler.
Why is k-fold cross-validation usually wrong for financial time series, and what do you use instead?
A gradient-boosted model predicts next-hour returns with an out-of-sample R² of 0.004. Is that useful?
You have 25 horses and a track that races five at a time, with no timer. What is the minimum number of races to find the fastest three?
One hundred prisoners each open 50 of 100 boxes searching for their own number; all must succeed or all fail. What strategy beats a vanishingly small success rate?
All questions here are written originally for this platform. We do not republish firms' proprietary interview questions.
Interview process by firm
The published shape of each firm's loop, compiled from public sources and candidate reports.