SIG · Online Assessment · Trader / Quant Researcher track

SIG Quantitative Assessment 2027: practice the Susquehanna online assessment, the 17-question math OA

SIG's first-round screen for Trader and Quant Researcher candidates. Three real-format questions are below — try them right now, no signup — then take the full timed simulation.

17questions
60minutes
Mettlplatform
Yescalculator
2019–26bank recurs
Q1–9 · fast tier, 1–2 min eachQ10–17 · deep tier, bank time for these

The 3 free questions below are pulled from across this curve.

Maintained by Stanley from the QuantVault team · last verified August 2026 against a candidate report from a summer-2026 sitting. How we keep this accurate. Jump to the practice questions ↓

What the test is really like

The Quantitative Evaluation opens SIG's funnel for the Trader / Quant Researcher track. It is a resume-screen filter, not a deep exam: the first ~9 questions are quick probability and expected-value calculations, the last ~8 step up to multi-stage problems — Bayes' rule, constraint logic, fair-coin gambler's ruin, and a lattice-path counting question that rewards knowing the dynamic-programming trick.

The single most useful fact from recent sittings: the question bank is largely fixed and has recurred for years — only the numbers change between sittings. That means drilling the known question families for speed is the whole strategy. At ~3.5 minutes per question, hesitation is what fails people, not difficulty.

Recent sitting report (candidate, Aug 2026): the live test closely matched our full simulation — no genuinely unseen problems — but a couple of items ran slightly heavier live: the canoe up/downstream question made the candidate solve for the paddling speed rather than handing it over, and the balance-mobile figure buried the same simple halving trick under many distractor branches (about 2 minutes just to find the branch that matters).

Sep 2026 sitting, captured question-for-question: a candidate's question-for-question report of the live Mettl test (on-screen title “The Susquehanna Problem Solving Assessment”) recovered all 17 questions with their figures. Every one is a number rotation of the families in the simulation on this page, but three prompts changed shape: the two-dice game now pays the lower roll and $0 on doubles, the bakery runs out of croissants (the 80% item), and the fake-news question asks for P(fake) rather than P(real). The screen itself is one section with a palette of 17 numbered tabs, an “Attempted: n/17” counter, a “Revisit Later” bookmark and autosave; answers go into Mettl’s own widgets — numerator / denominator blanks for fractions (“in simplest form”), a $-prefixed blank for the dice game, a %-suffixed blank for the bakery, hh:mm blanks for the canoe, a two-digit AB entry for the skyscraper grid, and a single five-option radio question (knights and knaves). We replicate that sitting word for word as SIG Problem Solving Assessment (Sep 2026 sitting), with a free 2-question example.

What the SIG online assessment screen looks like

The Quantitative Evaluation is hosted on Mettl (Mercer). Candidates report 17 questions in 60 minutes with a calculator allowed, and answers are typed as a number, a reduced fraction or a decimal. The first roughly 9 questions are quick probability and expected-value calculations and the last 8 are multi-stage problems, so the pace is about 3.5 minutes a question on average, with time banked in the fast tier for the deep tier.

The SIG Quantitative Assessment 2027 is the same screen in its 2027-cycle form. A full sitting leaked in mid-2026 with 16 of 17 problems recovered: thirteen were number rotations of the classic bank drilled in the simulation on this page, and three families were new to the rotation (a two-dice payoff-matrix expected value with a doubles penalty, a stars-and-bars style count, and an expected-consecutive-pairs linearity question). A separate 3-question, 12-minute replica runs exactly those three. For how Mettl compares with the SHL, HackerRank, Codility and Criteria screens other firms use, see which assessment platform each quant firm uses; for every firm's OA in one place, the OA practice hub.

The recurring families (from the actual bank)

FamilyBank exampleTier
Bayes posteriors try one ↓two-factory defect rates; "picked the loser"mid
Dice & game EV try one ↓3-dice expected valuefast
Gambler's ruin try one ↓bold-play variantdeep
Constraint logicround-table seating; knights & knavesmid
Binomial tailsbakery order fulfillmentmid
Lattice / DP countingno-3-in-a-row pathsdeep
Linearity of expectationexpected matching pairsmid
Relative speedcanoe up/downstreamfast
Diagram itemsbalance scale; skyscraper sudokumid
The bank barely changes between sittings — the question families stay fixed and only the numbers move, so drilling them for speed is the whole game. One thing worth knowing: a perfect 17/17 is still no guarantee, since the screen is one early filter among many. Treat it as necessary, never sufficient.

Try 3 real-format questions now

Same families as the real screen, original numbers. Enter a fraction like 3/8 or a decimal — a stopwatch starts when you begin typing.

Question 1 of 3 · Expected value0:00warm-up tier
You roll a fair six-sided die once. If it shows 1 or 2 you win $9; otherwise you lose $3. What is your expected profit, in dollars?
Solution. P(win) = 2/6 = 1/3, P(lose) = 2/3. EV = (1/3)(9) + (2/3)(−3) = 3 − 2 = $1. On the real screen this tier should take you under 60 seconds.
Question 2 of 3 · Bayes' rule0:00mid tier
A bag contains two coins: one fair, one double-headed. You draw one at random and flip it twice; both flips land heads. What is the probability you drew the double-headed coin?
Solution. P(HH | double) = 1, P(HH | fair) = 1/4. Posterior = (1 · ½) / (1 · ½ + ¼ · ½) = (1/2)/(5/8) = 4/5 = 0.8. SIG loves two-flip and three-flip Bayes setups; write the posterior as one fraction and you save a minute.
Question 3 of 3 · Gambler's ruin0:00late tier
You have $3 and your opponent has $5. You repeatedly bet $1 on fair coin flips until one of you is broke. What is the probability you end up with all $8?
Solution. For a fair game, P(reach N before 0 from k) = k/N = 3/8 = 0.375. This exact family — with shuffled bankrolls — appears in the late tier; memorize k/N for fair and the (1−(q/p)^k)/(1−(q/p)^N) form for biased.

Take the full 17-question timed simulation

The complete Quantitative Evaluation recreation: 17 questions, 60-minute clock, the real difficulty curve, auto-graded with worked solutions for every question. Built from the recurring bank.

Wondering what the offer is worth? See the SIG quant salary breakdown.

The full simulation is part of QuantVault Pro ($24/mo). The three questions above and the format guide are free, no account needed.

The other SIG online assessments

The 17-question math screen is the Trader / QR track. SIG runs different first-round OAs by track:

How to prepare (the honest version)

1Drill the families, not random mathEV of simple games, Bayes posteriors, gambler's ruin, constraint logic, lattice paths. The bank recurs; the families are known.
2Train pacing before content60 minutes / 17 questions ≈ 3.5 min each — and the early tier should take 1–2. Bank time early, spend it late.
3Take two full timed runsThe simulation grades you with worked solutions; your second run should feel mechanical.
4Aim near-perfectNo published cutoff — it's an early filter. Candidates who clear it report missing at most a couple.

How we keep this accurate

We track how candidates describe this screen over time and update the format, timing, and topic mix whenever the reports shift. Every practice question here is our own, written to match the question families and difficulty of the real screen, with answers derived and checked by code — not copied from anywhere. Where a detail can't be confirmed (like an exact passing cutoff), we say so rather than guess.

  • Sep 2026 — full sitting captured from a candidate’s report: all 17 questions and figures recovered and replicated word for word as a separate timed test; three prompt-shape changes noted (lower-roll dice game paying $0 on doubles, croissants as the scarce item, P(fake) asked directly); Mettl answer widgets documented (fraction blanks, $ / % blanks, hh:mm, two-digit AB, one radio question); bank twins re-checked against the live wording
  • Aug 2026 — candidate sitting report: bank confirmed question-for-question against our simulation; two items noted as running heavier live (solve-for-speed canoe variant, extra distractor branches on the balance mobile) — notes added to those solutions
  • Jun 2026 — format re-confirmed: 17 questions / 60 min, Mettl, calculator allowed
  • 2024–2025 — topic mix and difficulty curve re-checked against recent sittings
  • 2019–2023 — same question families seen throughout, numbers varying by sitting

FAQ

What is the SIG Quantitative Evaluation?

SIG's first-round online assessment for Trader and Quant Researcher candidates: 17 math questions in 60 minutes on Mettl, calculator allowed, covering probability, EV, Bayes, constraint logic, gambler's ruin, and lattice-path DP.

Is it the same as the "Susquehanna Problem Solving Assessment"?

Candidates use both names; the 17-question math screen described here is the quantitative first-round test for the trading/QR track. Other tracks get coding or finance assessments instead.

Can I use a calculator?

Yes — recent sittings on Mettl allow a calculator. Speed on setup, not arithmetic, is what's being tested.

What happens after the OA?

Passing is necessary but not sufficient — it’s an early filter. Next come interviews with probability, EV, and market-making games; see the full SIG interview funnel.

What is the SIG Quantitative Assessment 2027?

The 2027-cycle sitting of the same 17-question, 60-minute trader and quant researcher screen. Sixteen of its 17 problems were recovered: thirteen are number rotations of the classic bank and three families are new (a two-dice payoff-matrix expected value with a doubles penalty, a stars-and-bars style count, and an expected-consecutive-pairs linearity question). The 2027 set replica runs those three in 12 minutes; the full simulation on this page covers the other thirteen.

Is the Susquehanna online assessment the same as the SIG Quantitative Evaluation?

Yes. Susquehanna International Group is SIG, and candidates use Susquehanna online assessment, SIG OA, Problem Solving Assessment and Quantitative Evaluation for the same 17-question first-round math screen on the trader and quant researcher track. Other SIG tracks get different first-round tests: a proctored coding OA for software engineers, a CodeSignal test for strategy developers, and a finance OA for capital markets.

What does the SIG online assessment look like on screen?

One Mettl section titled “The Susquehanna Problem Solving Assessment” with a palette of 17 numbered tabs, an “Attempted: n/17” counter, a “Revisit Later” bookmark and autosave, so you can skip and come back. Sixteen answers are typed — fractions as numerator / denominator in simplest form, dollars and percentages into $- and %-labelled blanks with an explicit rounding rule, the canoe time as hh:mm, the skyscraper grid as a two-digit AB — and one question (knights and knaves) is a five-option radio choice. The Sep 2026 sitting replica reproduces all of that word for word.

What platform is the SIG online assessment on?

Mettl (Mercer). It is the only firm in our platform table on Mettl: 17 questions in 60 minutes, calculator allowed, answers entered as numbers or fractions, and a question bank that has recurred since 2019 with only the numbers changing.

Keep going