PRACTICE GUIDE JANE STREET
Jane Street Fermi Estimation Practice Test
Fermi estimation turns an unfamiliar quantity into a small calculation built from assumptions you can explain. Good estimation includes the reasoning, units and uncertainty, not just a guess near the reference.
Outcry offers Classic and Technical modes. Classic has ten questions at sixty seconds each; Technical has eight with question-specific clocks. These are Outcry settings, not verified timing for a particular Jane Street screen.
Outcry is not affiliated with Jane Street and has no access to their assessment content. This guide describes an assessment format that candidates report publicly; the questions here are generated by Outcry and are not Jane Street’s own.
What it screens
A quantitative trading firm active across equities, ETFs, bonds, options and currencies, known for OCaml and a strong internship pipeline.
- ✓Breaking a large unknown into quantities with explicit units
- ✓Constructing plausible bounds rather than false precision
- ✓Reading weighted averages, totals and density visuals
- ✓Estimating probability tails and bracketing roots
Where it sits at Jane Street
This page focuses on fermi estimation skills for Jane Street preparation. The linked exercise is built by Outcry and is not the firm’s assessment.
Check the role-specific invitation for required modules, timings and marking. These can differ by office and recruitment year.
The format
These are Outcry's practice-game settings, not verified figures for the firm's assessment.
| Questions | 10 Classic / 8 Technical |
|---|---|
| Time | Varies by mode and question |
| Per question | Classic: 60 seconds each; Technical: question-specific |
| Negative marking | No |
| Answer style | Enter a lower and an upper estimate |
| Where it sits | Outcry skill practice; follow your invitation for the firm's format |
What it tests, with a worked example
These worked examples are written by Outcry to explain the skills practised here. They are not the firm's questions.
Decompose a real-world total
Write a small model with units before choosing assumptions.
Example
Estimate daily coffee purchases in a town of 100,000 adults if 40% buy one coffee each day.
Answer About 40,000 coffees per day
100,000 adults × 0.4 purchases per adult per day = 40,000. This is an estimate conditional on the stated assumption, not a measured fact.
Build a plausible range
Propagate uncertainty in your assumptions into the final bounds.
Example
A venue admits 200–300 visitors per day. Each spends $10–$15. Estimate a daily revenue interval.
Answer $2,000 to $4,500
With these non-negative bounds, the minimum is 200 × 10 and maximum is 300 × 15. The interval covers the stated combinations; it is not a statistical confidence interval.
Use weighted averages
Weight each value by its frequency.
Example
A chart shows 10 sales at $20,000 and 40 at $30,000. Estimate the mean sale price.
Answer $28,000
Total value = 10 × 20,000 + 40 × 30,000 = 1,400,000. Divide by 50 sales. Averaging the two labels alone would incorrectly give 25,000.
Estimate a ledger total
Use a rounded anchor, then correct the largest deviations.
Example
A sheet has 12 entries near $80 each. Eight are exactly $80; four are $110. Estimate the total.
Answer $1,080
Start with 12 × 80 = 960. Four entries are 30 above the anchor, so add 120. The exact total here is 1,080.
Approximate a tail
State the model and distinguish an approximation from an exact probability.
Example
For 100 independent fair coin flips, estimate P(heads ≥ 60).
Answer About 3% using a normal approximation
The mean is 50 and standard deviation is 5. With continuity correction, z = (59.5 - 50)/5 = 1.9. The upper normal tail is about 2.9%; the exact binomial tail differs slightly.
Bracket a root
Test easy values to bound the answer before refining.
Example
Find a short interval containing the positive root of x³ + x = 32.
Answer Between 3.0 and 3.1
At 3, x³ + x = 30. At 3.1 it is 32.891. The function increases for positive x, so the unique positive root lies between these values.
What a good score looks like
Use accuracy and the reasoning behind each answer to judge progress. No verified firm pass mark is asserted here.
How to train for it
- 01Choose Classic for real-world decomposition, or Technical for charts, spreadsheets, probability bounds, roots and visual density.
- 02Write the model and units, then test a low and high set of assumptions. Avoid a narrow interval unsupported by your reasoning.
- 03After each answer, compare the reference and heuristic with your model; identify which assumption or approximation created the error.
TRAIN IT HERE
The drills that match each section
SIT THE FULL BATTERY
All the sections back to back on one clock, marked the way the real screen marks them, with a by-skill breakdown at the end. Included with any pass.
Also reported at Jane Street
Common questions
- What do I practise here?
- Estimate an unknown quantity by decomposition or rapid approximation, then submit a defensible lower and upper bound.
- How long is the practice?
- Outcry offers Classic and Technical modes. Classic has ten questions at sixty seconds each; Technical has eight with question-specific clocks. These are Outcry settings, not verified timing for a particular Optiver screen.
- How should I improve?
- After each answer, compare the reference and heuristic with your model; identify which assumption or approximation created the error.
- Where do I start?
- Use the Fermi Estimation game linked under Train it here after working through the examples.
Hiring now: current Jane Street quant openings, refreshed daily.