How Jane Street Interviews in 2026: Process, Questions & What They Score

• Classic coin-flip expectation problems, e.g. expected number of flips until two heads appear in a row.
• A 3x3x3 cube has all outer faces painted and is cut into 27 unit cubes; you pick one at random, drop it, and see the visible faces are unpainted: use conditional probability/Bayes' rule to reason about the hidden face(s) (a well-documented Jane Street Glassdoor question, seen in several phrasing variants).
• Two players each roll two dice; if any die value matches between the two players you win $1, otherwise you lose $1: would you play, and what's the expected value? (Confirmed multi-source Jane Street question; correct EV works out to +1/36 per game via 1 - 35/72 no-match probability.)
• You and a roommate host a party with 10 other couples; everyone (except you) is asked how many hands they shook, no one shakes their own partner's hand, and all reported counts are different: how many hands did your roommate shake?
• Divide 100 soldiers/units across 10 castles (worth 1-10 points) to maximize your expected score against opponents' hidden allocations (a Colonel Blotto-style resource-allocation game Jane Street has used both as a public puzzle and reportedly as an interview question).
• A treasure chest's true value is uniformly distributed between $0 and $1000 (unknown to you); you name a bid, and if your bid is at least the chest's true value you win it for that bid; a friend has agreed to then buy it from you for 1.5x its true value. What should you bid to maximize expected profit?
• A 20-sided die game: roll repeatedly over many rounds and either bank the current face value in dollars or reroll. What's your threshold strategy to maximize expected earnings?
• Dice-matchup variants comparing different-sided dice (e.g., a 12- or 20-sided die vs. a 30-sided die, sometimes with a reroll option) where the higher roll wins and the loser pays the winner, reasoning through expected value and the value of a reroll option.

• The interviewer reveals partial information about a hidden value (e.g., a card or number) and asks you to quote a two-sided bid/ask market, then reveals more information across rounds as you adjust.
• Estimate a real-world quantity (e.g., "how many gas stations are there in the US?") and make a market around it rather than giving a single point estimate.
• Play a sequential betting/casino-style game where your ending score in one round becomes your starting stake in the next, requiring running risk and bankroll management.
• Card-based market-making exercises where you must price a hidden card or hand and defend/adjust your market as opponents trade against you.
• Live, real-time versions of earlier phone-round probability puzzles, played out with actual chips/stakes rather than discussed theoretically, testing execution and composure, not just the math.

• Implement basic memoization of a function's results using a hash table, the documented first part of Jane Street's own public 'Memo' interview walkthrough (confirmed on blog.janestreet.com).
• Extend that memoization into a bounded/size-limited cache with FIFO eviction, aiming for O(1) operations, confirmed as part 2 of the same public 'Memo' example.
• Extend further into a full LRU (least-recently-used) cache with O(1) get/put, typically via a hash map plus doubly linked list, confirmed as part 3 of the same example.
• Design and implement a simplified limit order book supporting insert, cancel, and order-matching logic.
• Implement efficient streaming/running statistics (e.g., a moving average) without recomputing from scratch on each update.
• Parse and validate a structured data feed, handling malformed or edge-case input.
• Open-ended, multi-part problems where the interviewer layers new requirements onto the same evolving solution across the round rather than asking a fresh question each time, in a shared/collaborative code editor, in the language of your choice (Jane Street explicitly says it awards no bonus points for using OCaml).

• Given a sample dataset, how would you approach building a model to answer a specific practical question; Jane Street's own interviewing page describes ML/QR interviews as working through 'realistic modeling problems that give a flavor for the kinds of work we do every day.'
• Explain the bias-variance tradeoff and how you would detect/guard against overfitting.
• Multi-stage conditional-probability problems requiring Bayes' rule and the law of total probability (e.g., variants of the painted-cube problem).
• Basic statistics fundamentals: interpreting p-values, confidence intervals, or regression coefficients.
• Open-ended data-analysis / case-style prompts: 'here is some data: what do you notice, and what would you do next?'

• Why trading, and why Jane Street specifically?
• Tell me about a time you realized you were wrong about something and had to change your mind.
• Describe a mistake you made at work or school and how you handled/fixed it (directly echoes the firm's own trading-interview tip to 'correct your mistakes.')
• Tell me about a disagreement with a colleague, teammate, or professor and how it was resolved.
• What are your strengths and weaknesses?
• What do you do outside of work/school? (used to gauge genuine intellectual curiosity, not as small talk)

Jane Street's process has 5 stages, in order: Application / resume screen, Online assessment (role-dependent, not universal), Phone / video interview(s), Final round / onsite ("superday"), Decision.

Jane Street evaluates candidates against Jane Street does not brand one interview-wide framework the way some firms do (no "Leadership Principles"-style umbrella name), but it does publish an official, named, company-wide list plus two track-specific published lists, all verified directly against janestreet.com. The most complete and company-wide one is "Our Values," with exactly four named items, from janestreet.com/who-we-are/.: Ethics and Honesty: "No one at Jane Street will be asked to defend a position or state an opinion that they don't believe in... interactions both within and outside of our walls are conducted honestly and openly.", Responsible Risk Management: carefully studying how trading risks might be bigger or more interrelated than they appear, not taking excessive risk that others would bear the downside of, and running blameless postmortems when systems fail., Intellectual Humility: "Staying humble allows us to question our assumptions and to update our views based on new arguments... we talk openly about mistakes because that's the best way to learn from them.", Less hierarchy, more collaboration: minimal red tape, little emphasis on seniority, colleagues treated as "partners working together toward shared goals.".

Jane Street's question bank spans 5 categories: Probability, Expected Value & Brainteasers (Trading / Quant Research); Market-Making & Live Trading Games (Trading / Quant Research, onsite); Coding & Software Engineering; Quantitative Research & Statistics; Behavioral & Culture Fit.

This playbook is high confidence, compiled from 24 public sources, and last verified July 7, 2026. It describes one well-documented shape of Jane Street's interviews, not a guarantee of what any individual loop will look like.