Optiver’s first screen is 80 arithmetic questions in eight minutes, no calculator, and wrong answers subtract from your score. Most rejections in quant trading happen right there, before a human ever says hello. The filter fits the job: trading firms make money by computing expected value faster and more accurately than whoever is on the other side of the trade, and every round of the interview tests some version of that one skill.
The business, because the interview copies it
Quant trading firms trade financial instruments with models and software instead of gut feel, and the big names sort into a few overlapping camps. Market makers like Jane Street, Optiver, SIG, IMC, and Citadel Securities quote a two-sided price (a bid and an ask) on thousands of instruments all day, earn the spread between the two, and manage the risk of whatever inventory piles up. Systematic funds like Two Sigma and D.E. Shaw hunt statistical patterns and hold positions for minutes to months. High-frequency shops like Hudson River Trading, Jump, and Tower compete on latency, where an edge can live or die inside a microsecond. Prop firms trade the partners’ own capital; hedge funds manage outside money and charge fees for it.
A trader’s day at one of these firms is less shouting into phones and more supervising a fleet of models: watching positions, adjusting parameters, deciding when the model is wrong and a human should intervene, and pricing the weird one-off situations no model covers. Each interview round maps to a piece of that. Fast arithmetic is scanning a screen of positions. Probability puzzles are pricing. Betting games are risk management with something at stake.
Trader, researcher, or developer decides what you drill
Prep goes wrong most often because candidates study for the wrong track. Quant trader loops lean on mental math, expected value, and game rounds. Quantitative researcher loops go deeper on probability and statistics, add ML and data questions, and care far less about arithmetic speed; at the systematic funds a QR interview can feel like a stats qualifier crossed with a data challenge. Quant developer loops are software interviews with a latency accent: C++, data structures, concurrency, cache behavior, lock-free queues, and only light probability. Grind LeetCode for a trader seat at Optiver and you’ll fail the math test. Drill speed arithmetic for a Two Sigma research role and you’ve wasted a month.
The loop, round by round
Nearly every trading firm runs a version of the same funnel: an online assessment to filter, one to three technical phone rounds, then a final onsite that recruiters call a superday. The assessment is where styles diverge most. Optiver owns the famous 80-in-8 arithmetic test, and Flow Traders runs the same format, mostly fractions, percentages, and conversion problems under a clock that never feels long enough. SIG’s Quantitative Evaluation mixes probability with logic puzzles, grid path counting and small systems of equations among them. Jane Street skips timed arithmetic entirely and opens with a probability conversation on the phone; the firm has said repeatedly it cares more about numeracy and sound reasoning than raw speed. Our Jane Street guide covers that loop in detail.
Phone rounds mean brainteasers and expected-value problems worked out loud with a trader listening. The final round adds the games: market-making exercises, group trading simulations at Optiver, betting sequences where an interviewer offers wagers and watches how you size them, poker hands at SIG. The games carry heavy weight because they’re the closest simulation of the actual job a firm can run in a conference room.
| Firm | Trading style | First technical filter | What the trader interview leans on |
|---|---|---|---|
| Jane Street | Market making (ETFs, options, bonds) | Probability phone screen; no timed arithmetic test | Probability intuition, betting and estimation games, reasoning out loud |
| Optiver | Options market making | 80 arithmetic questions in 8 minutes, errors penalized | Mental math speed, market-making games, confidence calibration |
| SIG (Susquehanna) | Options market making | Quantitative Evaluation (probability and logic puzzles) | Decision theory, poker-style betting, expected value |
| Citadel Securities | Market making at scale | Online math and coding assessment | Probability plus more coding than most trader loops |
| Hudson River Trading | Automated / high frequency | Coding and probability assessment | Algorithms and probability; systems depth for dev-track roles |
| Two Sigma | Systematic hedge fund | Coding or stats assessment | Statistics, ML, data reasoning; little mental-math speed |
| D.E. Shaw | Systematic and discretionary hybrid | Technical phone screens | Hard probability brainteasers mixed with CS |
| Flow Traders | ETF market making | 80 arithmetic questions in 8 minutes | Raw calculation speed, then trading games |
Questions phrased the way traders ask them
- “You roll a fair die and get paid its face value in dollars. You may re-roll up to two times, but you keep only the final roll. What’s this game worth?”
- “Make me a market on the number of commercial flights departing the U.S. each day.”
- “I’ll flip a fair coin 100 times and pay you a dollar per head. What would you pay to play?”
- “I flip through a shuffled deck one card at a time. Say stop whenever you like, and you win if the next card is red. What’s your strategy?”
- “Quick: 17 times 24.” (dropped mid-conversation, with a clock running in the interviewer’s head)
The die game deserves a walkthrough because it shows the reasoning style firms want to hear. A single roll is worth 3.5, the average face. With one re-roll left, keep anything above 3.5, so you hold a 4, 5, or 6: half the time you keep an average of 5, half the time you re-roll into a fresh 3.5, and the game is worth 4.25. With two re-rolls, the first decision compares against 4.25 rather than 3.5, so now you only hold a 5 or 6: a third of the time you keep an average of 5.5, two thirds of the time you continue into the 4.25 game, about 4.67 in total. Working backward from the last decision is dynamic programming said out loud, and interviewers notice candidates who find that structure instead of flailing at cases.
The deck question is a trap pointed the other way. No stopping rule beats 50 percent, and there’s a clean argument for it: at any moment your chance of winning equals the fraction of red cards left in the deck, and that fraction has no drift you can exploit. Strong candidates test a two-card case, smell the symmetry, and commit to the plain answer rather than inventing a clever-sounding rule. Recognizing that there’s no edge is a trading skill too.
The coin-flip question separates people who know one formula from people who think like traders. Expected value is $50, so that’s the risk-neutral price. A good answer goes one step further without being asked: the standard deviation is $5, so even paying $49 loses money nearly 40 percent of the time, and how far below $50 you’d actually pay depends on your bankroll. Interviewers light up when a candidate connects price to variance and bet size unprompted.
Make me a market, and what good looks like
The signature round starts with an unknowable quantity, flights per day or piano tuners in Chicago, and you quote a bid and an ask. Quote 3,000 bid at 40,000 ask and you’ve said nothing. Quote 24,000 at 25,000 and you’d better have reasoning behind that confidence. Then the interviewer trades against you. If they lift your offer, suspect they know the number is higher, move your market up, and widen it. Keep quoting the same prices as information arrives and you fail; refuse to trade at all and you also fail. The round scores whether you update on evidence and whether your stated uncertainty matches your actual uncertainty.
Calibration gets tested directly too. Optiver likes 90 percent confidence-interval quizzes: give intervals for ten trivia quantities, and a well-calibrated candidate misses about one. Miss five and your intervals were far too narrow, which reads as overconfidence, the one trait a trading firm cannot hire because overconfident traders lose money at size. A wide, reasoned interval beats a confident wrong number every single time.
The other consistent signal is what happens when you get pushed. Interviewers at these firms argue against your answer whether you’re right or wrong. Fold instantly and you look like the market will run you over; dig in against clear logic and you look worse. Jane Street’s own advice says they’d rather watch you work through a wrong approach, catch the mistake, and correct course than sit silent until you’re certain. Every firm on the list scores some version of that.
Timing, prep, and how learnable this is
The calendar runs absurdly early. Jane Street, Citadel, Hudson River Trading, Two Sigma, D.E. Shaw, Optiver, and SIG all open summer 2027 internship applications in August 2026, a full year out, and the strongest firms fill most full-time seats from their intern classes. If you want a trading seat, the application window sits a year before the desk, and showing up in spring means waiting a cycle.
The prep itself is unusually drillable. Ten minutes a day on an arithmetic trainer like Zetamac moves your speed more than any other habit. Xinfeng Zhou’s “A Practical Guide to Quantitative Finance Interviews”, the book everyone means when they say green book, covers the probability canon. Practice making markets on random quantities with a friend who trades against you and bluffs sometimes. First-year total comp at the top shops sits in the mid six figures in public reports, with the caveats that numbers move yearly, vary hard by firm, and lean on a bonus nobody guarantees. For that payoff this is the cheapest prep in tech: a few months of drills, no portfolio, no take-homes, and a process that tells you exactly what it wants. Expected value, computed quickly, defended calmly, and revised the moment the evidence says you’re wrong.
