Optiver Interview Guide (2026)

1Snapshot, read this first#

The one thing to understand about Optiver:

• They are not hiring you for what you already know. They are testing how fast and how clearly your brain works when the clock is running and you are uncomfortable.
• The content is almost a pretext. A dice problem is not about dice.
• It is about whether you can define a random variable, build an expected value, and stay organized while someone watches.
• Treat every question as a window into your process, because that is exactly how they read it.

2Getting the call (the part most guides skip)#

Optiver gets flooded with applications, especially for new grad and intern spots. The filter is brutal and mostly automatic early on.

• For the most sought-after new grad trader and quant seats, a single Amsterdam or Chicago opening can pull thousands of applicants.
• The online assessment does most of the cutting before a human ever reads your resume.
• So the moves below are about getting a human to read you at all, then not wasting that read.

Target the pipeline, not the posting.

• Most new grads come in through campus recruiting, their internship (the single best door), or structured early-careers programs tied to Amsterdam, Chicago, Austin, London, Sydney, Singapore.
• The internship is the strongest door by a wide margin: a large share of new grad seats go to people who already interned and got a return offer.
• If your school is on their list, the campus channel beats the general portal, because campus applicants go into a separate, smaller, human-reviewed pool.

Referrals matter, but less than at a FAANG.

• A referral gets a human to glance at your resume. It does not get you past the OA.
• At Optiver the OA is the real gate, so spend your energy there.
• Still, if you know someone inside, ask. A referrer who vouches for your composure under pressure is worth more than one who just forwards your PDF.

Make your resume readable in 8 seconds. Recruiters here skim for signal: competitive programming rating, math olympiad, poker/trading/markets interest, systems projects, measured impact. A bullet that works:

Built a lock-free order-book matching engine in C++; cut median message handling latency from 4.2us to 1.1us by replacing a mutex queue with a single-producer ring buffer, benchmarked over 10M synthetic orders.

Why that line lands:

• It has a number, a technique, a before/after, and a verification method.
• The verification method is the part most candidates skip, and the part a trading firm respects most.
• Quoting a latency number without saying how you measured it is exactly the kind of unverified claim they trade against all day.

ATS keywords that help (only if true):

• C++, low latency, concurrency, lock-free, data structures, algorithms, probability, statistics, Python, Linux, networking, Codeforces/LeetCode rating, options, market making.
• Do not keyword-stuff. They read these resumes closely, and a buzzword you cannot defend in the phone screen is worse than leaving it off.

If you have a competitive rating, put it near the top.

• For this crowd a 1900+ CF rating (Candidate Master and up) or a strong ICPC/IOI/Putnam result is a louder signal than a 3.9 GPA.
• A GPA says you studied. A 2100 CF rating says you solve hard problems fast under a clock, which is the one thing this whole loop measures.

3The loop at a glance#

How to read this table:

• Not every candidate sees every round. Trader pipelines lean math/game heavy; SWE pipelines lean coding/systems heavy. Many people see a mix.
• The super-day is usually a single intense block: three to five rounds back to back in a morning or afternoon.
• There is often a short break and sometimes lunch with current traders or engineers that is more evaluative than it looks.
• Treat every conversation in the building as part of the loop.

4The coding or technical rounds#

The honest truth: Optiver's coding bar is high but it is not exotic. They reuse the same families of problems, and they care more about how you get there than whether you memorized the trick.

The patterns that keep showing up:

• Arrays, hashing, two pointers, sliding window.
• Sorting and binary search, including binary search on the answer.
• Stacks and monotonic stacks (order books and price sequences love these, and "next greater element" style problems map directly onto market data).
• Heaps / priority queues (think "k best quotes" or merging streams).
• Graphs: BFS, DFS, shortest path, topological sort.
• Dynamic programming, usually medium not nightmare.
• Simulation problems, often framed around markets, orders, or sequences. A common shape: replay a stream of add/cancel/trade messages and answer a query about the book state. Get comfortable building a small matching engine on the fly.

On difficulty:

• Think LeetCode medium to the easier end of hard, but under real time pressure and with a sharp interviewer poking at your complexity.
• The OA especially rewards raw speed: you may get 3 to 4 problems in 60 to 90 minutes, so partial-but-fast often beats perfect-but-slow.
• My rough bar: a medium coded, tested, and explained in under 20 minutes end to end. If you cannot solve a clean medium in 15 minutes, drill until you can.

For low-latency SWE specifically, expect C++ depth:

• Move semantics, where allocations happen, cache behavior, false sharing, std::vector vs raw buffers.
• When a branch or a mispredict costs you, lock-free vs locking, memory ordering at a basic level (relaxed vs acquire/release and why it matters).
• They will ask "what actually happens in memory here" and "how would you make this faster."
• A real example they like: given a hot loop, explain why std::vector<std::shared_ptr<T>> thrashes cache and what you would use instead.
• Know your language and the machine under it, not just your algorithms.

How to actually win the round:

1. Clarify first. Restate the problem. Ask about input size, ranges, duplicates, edge cases. Thirty seconds here saves you a wrong solution. If n fits in 10^5, you have room for O(n log n). If it is 10^9, they are hinting at math or binary search, not iteration. The constraint is the clue.
2. Talk the whole time. Silence is the number one killer. Say what you are checking: "I am scanning left to right keeping the max-so-far, because I only need the best window ending here." They are grading your reasoning, and they cannot grade silence.
3. State complexity before and after. "This brute force is O(n squared), I think I can get O(n) with a hashmap." Then deliver. Naming the target complexity before you code shows you have a plan, not a hope.
4. Test your code out loud. Pick a small input, walk it. Catch your own off-by-one before they do. A trading firm cares a lot about the person who finds their own bug.
5. Write clean code. Good names, no spaghetti. These people read production code daily and they notice. Single-letter variables in a 40-line function read as sloppy here, not clever.

5Probability, brain-teasers, and the market-making game#

This is the Optiver signature, and for trader roles it is the whole ballgame. Even SWE candidates sometimes get a probability question.

Mental math test.

• Usually around 80 arithmetic questions in about 8 minutes: decimals, fractions, negatives, percentages, multi-step. No calculator.
• That is roughly 6 seconds per question with no slack, so hesitation is death.
• The trick is not genius, it is drill. Practice until two-digit multiplication, fraction-to-decimal (know 1/7 = 0.142857, 3/8 = 0.375, 5/6 = 0.833 cold), and quick percentage work are automatic.
• Use "zetamac" style arithmetic drills daily, set to the real mix: addition, subtraction, multiplication up to two digits, and division.
• Target comfortably past the threshold, which tends to sit around the high 50s to 60s correct out of 80, though they do not publish it. Build a 10-point buffer so a shaky day still clears.

Probability and brain-teasers. Expect:

• Expected value problems. Dice, coins, cards, drawing until some condition. A staple: "expected number of rolls to see a 6," answer 6, and they will push you to derive it with the recurrence E = 1 + (5/6)E.
• Conditional probability and Bayes-style updating. "You see one positive test, what is the real probability you are sick," with a low base rate that trips up gut instinct.
• Combinatorics and counting. Clean handling of choose vs permute, and overcounting.
• Fair-value / betting puzzles: "I offer you this game, what would you pay to play?" The answer is the EV, and then they ask what spread you would quote around it.
• Classic teasers: 100 prisoners and the light switch, two eggs and a 100-floor building, pirates splitting gold, the Monty Hall family. Know the canonical solutions and, more importantly, the reasoning so you can rebuild one you half-forgot.

They care about your framework, not your gut:

• Define the random variable, write the EV as a sum, check the edge cases, give a number, then sanity-check it.
• Saying "my gut says around 3.5, let me verify" and then verifying is exactly the behavior they want.
• When you get stuck, say the smaller version out loud: "let me solve it for 2 dice first, then generalize." That move alone keeps you in the game.

The market-making game. This is where traders are made or broken. The setup varies but the shape is constant: you quote a two-sided price (a bid and an ask) on something uncertain, the interviewer trades against you, and new information arrives that should move your price. They are watching:

• Do you set a sensible spread around your fair value? Too tight and you get picked off, too wide and you never trade. Both are penalized.
• Do you update correctly when news arrives, or do you anchor stubbornly?
• Do you manage risk and position, or drift into a huge one-sided bet? If you are already long 20 and they keep hitting your bid, your quote should move, not stay frozen.
• Do you stay calm and keep talking when you are losing?

A simple example of how one plays out:

Interviewer: I'll make a market on the number of people in this building. Give me your market.

Candidate: Fair value feels like 400, I'll quote 380 at 420.

Interviewer: It's a 40-floor tower.

Candidate: That's bigger than I assumed, so I'll shift up and tighten a touch: 440 at 470.

What that exchange is really testing:

• The exact number matters far less than disciplined updating and sane spreads.
• If the interviewer lifts your offer twice in a row, they are telling you your price is too low.
• The strong candidate notices and raises; the weak one keeps selling at the same level and bleeds.
• Do not try to be perfect. Be consistent, quick, and honest about your uncertainty. Traders who freeze or refuse to quote get cut, because in the real seat not quoting is itself a losing decision.

6Behavioral and culture fit#

Optiver's culture is genuinely intense and genuinely collaborative at the same time.

• The values they actually live: ownership, curiosity, intellectual honesty, teamwork, and improving things relentlessly.
• They want people who are hungry to understand how things work and who can admit when they are wrong without ego.
• The intellectual honesty piece is not a poster slogan. In the probability and market-making rounds they will push back on a correct answer just to see if you fold.
• Holding your ground with a clear reason, or cleanly saying "you are right, I missed that," both score. Getting flustered loses.

Prepare 4 to 6 STAR stories:

• A time you were wrong and fixed it, a time you made a hard call under uncertainty, a time you improved something nobody asked you to, a conflict you handled.
• Short, specific, with a number in the result.

One worked STAR example:

Situation: Our team's nightly risk report ran for 50 minutes and kept missing the pre-market window. Task: I owned getting it reliably done before 6am without buying new hardware. Action: I profiled it, found 70% of the time was a single unindexed join re-reading the same table in a loop. I pulled that read out, cached it in memory, and parallelized the per-account calc across cores. Result: Runtime dropped from 50 minutes to 6, and we hit the window every day for the next six months. I also wrote a one-page runbook so I wasn't the single point of failure.

Why that story works:

• It shows ownership, curiosity, measurement, and that you do not hoard knowledge.
• There is a real number in the result and a measurement step in the action.
• "I made it faster" is a non-answer here; "50 minutes to 6" is the answer.

On "why Optiver":

• Have a real reason tied to markets or systems: pricing under uncertainty, the engineering challenge of sub-microsecond trading, the fact that feedback is immediate and measured in P&L.
• "The pay" is true for everyone and signals nothing.

7Your prep plan#

4-week plan

• Week 1: 40+ LeetCode mediums across the pattern list. Start daily arithmetic drills, 15 min. Log your medium solve times so you can see them drop.
• Week 2: Probability and EV every day. Work through a problem set of brain-teasers. Keep arithmetic drills going, push for speed and track your score out of 80.
• Week 3: Timed mock coding, out loud. For SWE, drill C++ internals and low-latency concepts. Start practicing market-making games with a friend who keeps throwing new info at you.
• Week 4: Full mock loops. Refine STAR stories. Tighten arithmetic to beat the clock comfortably. Rest the day before.

2-week plan

• Days 1 to 5: Mediums daily + arithmetic drills + 1 probability set per day.
• Days 6 to 9: Timed mocks, market-making practice, C++ review for SWE.
• Days 10 to 12: Full probability + brain-teaser review, STAR stories.
• Days 13 to 14: Light mocks, arithmetic speed, sleep.

2-day crash plan

• Day 1: 10 mediums fast, 1 hour arithmetic drilling, review EV and conditional probability, 1 market-making mock.
• Day 2: 5 more mediums, re-drill arithmetic, skim classic brain-teasers, write 3 STAR stories, sleep early.

If you only have time for one thing:

• Make it the arithmetic drill. It is the most trainable round and the one where unprepared smart people lose the most points.

8Why people get rejected at Optiver#

1. They freeze on mental math. Smart people who never drilled arithmetic bomb the 8-minute test. It is pure practice and they did not do it. Getting 35 of 80 when the bar sits near 60 ends it right there.
2. They code in silence. Correct answer, zero narration, interviewer cannot tell how they think. Rejected.
3. They anchor in the market-making game. News arrives, their price does not move, they take a dumb one-sided position. That is the exact opposite of a trader.
4. They cannot handle being wrong. Optiver pushes back hard on purpose. Candidates who get defensive or crumble fail the honesty test.
5. Sloppy probability. They eyeball an EV instead of defining the variable and summing it, get the wrong number, and cannot explain the gap.
6. No genuine curiosity. "Why Optiver?" answered with "high pay" and nothing about markets, systems, or how things work. That reads as a flight risk.
7. Slow on the OA. Strong candidates who solve 2 of 4 OA problems get cut because speed is the whole point of that stage. The OA is a filter, not a conversation, so banked speed is everything.

9The offer, comp and negotiation#

Optiver pays at the very top of the market, and for new grads the bonus and sign-on often dwarf the base. Rough ranges:

• US new grad trader: base often ~$150k to $200k, plus a sizable sign-on and a bonus that regularly pushes total first-year comp into the $300k+ range.
• US SWE: base ~$150k to $200k+ with strong bonus and sign-on; senior low-latency specialists earn well beyond that.
• London: trader base often ~£80k to £120k+ with a large bonus and sign-on.
• APAC (Sydney/Singapore): very strong local comp through those hubs; India direct hiring is thinner and usually routed through these offices.

A key structural point:

• Unlike a FAANG offer, there is usually no equity.
• The lever is variable comp, the bonus, which is where the real money sits and the least visible number going in.
• Ask how it is determined (firm performance, desk performance, your performance) and over what window it is assessed.
• That answer tells you more about your real earnings than the base does.

Negotiation tips:

1. Use competing offers honestly. Optiver responds to real market data, especially rival prop/quant offers. A genuine Jane Street or Citadel offer moves numbers. A made-up one does not survive one follow-up question, and these are people whose job is reading bluffs.
2. Negotiate the sign-on and the structure, not just base. The variable comp is the big lever here. Ask how the bonus is determined and whether the sign-on is flexible. Sign-on is often the easiest number for them to move.
3. Be direct and unemotional. This is a firm of traders. A calm, specific "I have X, can you get closer" lands far better than a long emotional pitch. They respect clean negotiation and they close fast when the numbers are real.

When you are ready, map these rounds onto your actual calendar and build a dated plan backward from your interview date so nothing gets crammed at the end.