insideout Get Pro
Quant Researcher

Quant Researcher

The night before

Read it tonight, then once more on the way in.

Back to the rounds

1The day

  1. Online quizA timed online test of probability, statistics and maths, at the level of a strong maths degree.
  2. Maths interviewA harder problem worked through with a researcher, who watches how you get unstuck more than whether you finish.
  3. The research caseYou're shown a signal or a backtest and asked what's wrong with it: data from the future, companies missing from the history, too many ideas tried, or a test that ignores time.
  4. Modelling caseYou're given a prediction problem with no single right answer and asked how you'd model it, out loud.
  5. Coding with dataClean a dataset, compute some statistics, maybe write a small backtest, in the language you know best.
  6. Meeting the teamQuestions about research you've done and why you made each choice, then meeting the team you'd join.

2How to answer

Work through a modelling case30 min

  1. The exact target5 min
  2. Inputs known at the time6 min
  3. A baseline5 min
  4. An honest test8 min
  5. What would make you drop it6 min

Solve a probability problem out loud15 min

  1. Restate1 min
  2. Try a small case3 min
  3. Choose the method2 min
  4. Compute5 min
  5. Sanity check2 min
  6. Take the hint and the follow-up2 min

Tell a story from your work2 min

  1. The situation in one line15 sec
  2. What you did60 sec
  3. The result, with a number30 sec
  4. What you'd change15 sec

Walk through your project10 min

  1. The problem2 min
  2. Your choice and its alternative3 min
  3. What broke2 min
  4. The result2 min
  5. What's next1 min

3Numbers to have ready

Say each answer out loud, then open the fold.

  • The t-statistic of a strategy with an annual Sharpe ratio of 1 over four years.

    Sharpe × √years = 1 × 2 = 2.0, so p is about 0.046. Tests and p-values

  • Twenty useless ideas tested at 5%: the chance one of them looks real.

    1 − 0.95²⁰, about 64%, and one false winner on average. Multiple testing

  • The Bonferroni line for those twenty tests.

    p below 0.05 / 20 = 0.0025, or t beyond about 3.02. Multiple testing

  • The standard error of a 10-day average for a stock that moves 2% a day.

    2% / √10, about 0.63%. Four times the days halves it. Samples and estimates

  • 25 days average 0.30% with a standard deviation of 2.0%. The 95% interval.

    0.30 ± 1.96 × 2.0 / 5: from −0.48% to +1.08%. Zero is inside it. Confidence intervals

  • Up 50%, then down 50%. What's left?

    1.5 × 0.5 = 0.75, a loss of 25%. Prices and returns

4Traps

  • I'd train a deep network on the whole history.It skips the baseline, so you can't tell whether it learned anything. And daily data gives a few thousand rows, little for a model with many weights.
  • It's 36.A bare number, right or wrong, shows nothing. They're grading the route, so say it as you go.
  • We decided, we built it, and we fixed it.They're hiring you. With "we" all the way through, they can't tell what you did.
  • The stack was Python, Postgres, Airflow and Docker.A list of tools isn't a decision. Say what you chose, what you turned down, and why.

5On the day

  • Ask which AI tools, if any, you may use in each round.
  • Say your plan before you solve, so they can steer you early.
  • Stuck for a minute? Say what you've tried and ask for a hint. Taking one well counts.
  • Check units and sizes out loud before you give a number.
  • Before any model, say exactly what you'd predict and when each input was known.