Optiver Online Assessment

The Optiver online assessment is a speed test, not a coding marathon. For the Quant Trader and Quant Researcher tracks it is a timed mental-math and probability battery, where finishing accurately under the clock matters more than knowing any one formula.

Software Engineers take a different shape entirely: long object-oriented simulation coding problems plus a CS quiz. This page covers both tracks.

Last updated May 29, 2026 · maintained by Stanley on the QuantVault team — aggregated from candidate write-ups across cycles 2019–2026, verified against public format clones. Figures are candidate-reported, not employer-confirmed.

Format
Mental-math & reasoning battery (SWE: coding)
Length
Role-dependent (see tracks)
Platform
CodeSignal; QR adds HackerRank
Difficulty
Very hard — auto-rejects on submission

Where the OA sits: it is the first gate in Optiver's funnel; clearing it leads to the live betting / market-making game, then HR and track-specific rounds. See the full Optiver interview funnel ›

Format, timing & platform

For quant roles the Optiver online assessment is a multi-section battery on CodeSignal, about three hours total and often spread across days. It is deliberately long and anti-GPT, and frequently auto-rejects the moment you submit. The famous 80-in-8 mental-arithmetic screen — 80 questions in 8 minutes, no calculator — is a separate assessment from this battery; together they form the front gate.

The battery has four section types: Number Logic sequences (~15 questions / 25 min), Beat the Odds rapid probability (~30 questions / 45 min, ~90s each), and the Zap-N reaction and memory mini-games (45–60 min). The probability and sequence modules score +1 correct, −1 wrong, 0 skip with no going back — so a confident skip beats a blind guess.

What's confirmed vs. what varies: the section types, the +1/−1 scoring and the no-calculator rule hold across years. Section count and timing drift by cycle and role — recent Quant Trader / PhD runs show three Zap-N games rather than nine, and a 2026 variant adds a "Likelihood Test" (rank three events by probability from a chart). Treat minute counts as representative, not contractual.

What's tested

Most items are not meant to be solved exactly — you estimate the magnitude and snap to the closest option. The table gives a representative one-line stem per area so you can feel the difficulty, not just read the name.

Optiver OA — tested areas, representative stems and what makes each hard (quant tracks)
AreaRepresentative stem & why it's hardDifficulty
Mental arithmetic (80-in-8) 66 × ? = 138.6 — find the missing operand. ~6 seconds per item with a −1 penalty; the missing-operand shape is a division in disguise. very hard
Number Logic (sequences) Next term of 2, 5, 12, 29, 70, … (here a(n)=2·a(n−1)+a(n−2)). No calculator; the last few sequences are near-impossible, and wrong answers cost a point. hard
Beat the Odds (probability/EV) Closest value: expected return time to the start on a 6-node cycle. ~90s each, five options, pick the closest — recurring families must be reflexes. hard
Zap-N reaction games Code Compare: match a 7–10 char string from 5 choices, speeding to ~0.5s. Dual-task load and short-term recall under time pressure; nerve, not strategy. medium
Coding — SWE simulation Merge two redundant, reordered packet streams into one de-duplicated sequence. Long OOD problems where corner cases and the final method eat the clock. hard

How the OA differs by track

Timing and platform are role-dependent, so there is no single top-line number. The quant tracks share the mental-math battery; the SWE track is a different assessment built around long coding problems.

Quant Trader / Researcher track

Sections
Number Logic, Beat the Odds, Zap-N games, plus the separate 80-in-8 arithmetic screen
Length & platform
~3 hr on CodeSignal, often across days; no calculator
What to expect
Estimation and probability reflexes under a clock; QR adds a 90-min, 3-problem HackerRank algo round

Software Engineer / Quant Dev track

Sections
One or two long object-oriented simulation problems, then a ~20-question CS-knowledge quiz
Length & platform
75–90 min per problem on CodeSignal
What to expect
Implement a small class, method by method; corner cases and the final method are where time runs out

Worked examples

Three short play-throughs in the OA's real forms — each shows the recognise-then-read-off rhythm the speed sections reward.

Beat the Odds — return time on a cycle

Takeaway: for a symmetric walk on a ring of n nodes, the expected return time to the start is exactly n — instant once you recognise the family.

Q. A token does a left/right random walk on 6 nodes in a circle. Expected steps to first return to the start? Closest of 3, 6, 9, 12.

1. Recognise the family. “Return on an n-cycle” is a recurring Beat-the-Odds shape, not a fresh problem.

2. Recall the result. Mean return time to any node equals n.

3. Read it off & sanity-check. With \(n = 6\), the answer is 6. The move: name the family, then read the closed form — the trap is grinding a recursion when the family gives it instantly.

Number Logic — the doubling recurrence

Takeaway: when first differences show no pattern, test the two-term recurrence a(n) = 2·a(n−1) + a(n−2) — a recurring Optiver trap.

Q. What comes next in 2, 5, 12, 29, 70, … ?

1. Differences fail. Gaps 3, 7, 17, 41 — not arithmetic or geometric.

2. Test the recurrence. \(a_n = 2a_{n-1} + a_{n-2}\): \(2\cdot5+2=12\) ✓, \(2\cdot12+5=29\) ✓, \(2\cdot29+12=70\) ✓.

3. Apply it. Next is \(2\cdot70 + 29 = 169\). The move: bank gettable sequences early; skip the near-impossible last few.

SWE coding — merging two redundant streams

Takeaway: the OOD problems are a clean class with a bounded buffer and correct ordering — not a clever algorithm. Model the state, then handle duplicates and reordering as packets arrive.

Q. Redundant-channel merge: two streams carry the same sequence-numbered packets, possibly reordered or duplicated; emit one in-order, de-duplicated stream.

1. Track the next expected id and a small map of buffered out-of-order packets keyed by sequence number.

2. On arrival, drop if the id is already emitted or buffered (handles A/B redundancy); else buffer it.

3. Flush greedily. While the next expected id is buffered, emit and advance — keeping the buffer bounded. The grade is in the corner cases: late duplicates and gaps that never fill — test your own driver, as in the real editor.

Warning: Time traps to avoid
  • Don't guess into the −1. A wrong answer is a two-point swing versus a skip — answer only with a real estimate.
  • Don't over-invest in one item. Number Logic's final sequences are near-impossible; banking the easy ~15 is the safe score.
  • On coding, budget for the end. Corner cases and the final method are where most SWE candidates run out of time.

How to prepare

Prep is about reflexes, not coverage — work this order:

  1. Rehearse the 80-in-8 in the exact format. Drill times tables to 19 and the common fraction–decimal conversions, and practise the missing-operand shape until reframing it is automatic.
  2. Turn the Beat the Odds families into 30-second reflexes. The recurring shapes — cycle return times, gambler's-ruin ratios, symmetry “remove pairs” expectations, coupon-collector sums — repeat across years; drill each until it is instant.
  3. Learn Number Logic's recurring tricks. Alternating odd/even sub-sequences, the two-term recurrence, product-of-previous and difference ladders cover most items; simplify any disguised constant first.
  4. Practise the Zap-N reaction games cold. They reward calm and speed, not strategy — rehearse short-term recall, dual-task arithmetic, and a consistent risk policy on the balloon game.
  5. If you're on the SWE track, grind long OOD simulations under a clock. Build a small class end to end in 75–90 minutes and force yourself to finish the last method with corner cases handled.

Free practice for the Optiver OA

Quant tracks. Each link opens a self-contained trainer in the real format — no signup for the interactive screens.

Free coding problem (SWE track)

The SWE OA is built around long object-oriented simulations like the sketch above. This one is free to work in full, with a complete solution:

Optiver OA — FAQ

What is the Optiver online assessment?

It is a timed mental-math and reasoning battery, not a single coding test. Quant candidates get Number Logic, Beat the Odds and the Zap-N games plus the separate 80-in-8 arithmetic screen; Software Engineers instead get long object-oriented simulation problems and a CS-knowledge quiz.

How long is the Optiver OA and what platform is it on?

It is role-dependent. The quant battery runs about three hours on CodeSignal, often across days; the SWE track is 75–90 minutes per problem on CodeSignal, and the QR add-on is a separate 90-minute HackerRank round. No calculator on the math modules.

What is the pass rate for the Optiver online assessment?

There is no official figure — it is a heavy filter that often auto-rejects on submission. The clearest public benchmark is the 80-in-8, where candidate reports put the pass line near 55 net points and ~70-plus as competitive. Treat it as a difficulty signal, not a published rate.

How do I prepare for the Optiver online assessment?

Build speed reflexes, not breadth. Rehearse the 80-in-8 in the exact format, make the recurring Beat the Odds families instant, learn Number Logic's sequence tricks, and practise the Zap-N games cold — or, on the SWE track, grind long simulations under a 75–90 minute clock.