More quant candidacies end at the online assessment than at any human interview. The assessment is the first filter, it often arrives within days of applying, and most candidates walk into it having prepared for interviews instead. This guide catalogs the assessment formats trading firms are using in the 2026-27 cycle, as candidates report them, and gives the training method for each family.
One framing note and one disclaimer. The framing: formats cluster. Dozens of firms, but only about five families of test, so preparing by family beats preparing by firm. The disclaimer: QuantPit is independent and unaffiliated with every firm named here; formats are described from consistent public candidate reports, they shift between cycles, roles, and regions, and your invite email is always the final word on what you will face.
Family 1: speed arithmetic
The format. Many mental arithmetic questions against a short clock, the archetype being the screen candidates call the Optiver 80 in 8: eighty questions, eight minutes, commonly reported scoring of +1 correct, -2 wrong, skips free. Question content is deliberately elementary, integer operations, decimals, fractions, percentages; the test is speed and accuracy under a penalty rule, not mathematical depth.
What it filters for. Calibration. The -2 makes guessing ruinous, so the test rewards knowing which answers you are certain of, which is a trading skill wearing an arithmetic costume.
How to train. Daily short timed drills, technique work on complements and left-to-right arithmetic from our mental math tricks guide, and full-format runs. Our free 80 in 8 benchmark reproduces the format and scoring and returns your percentile against other candidates; the dedicated 80 in 8 guide has the pacing math and a two-week plan.
Family 2: sequences and pattern tests
The format. A stream of number sequences, find the next term, under aggressive time limits, the archetype being the test candidates associate with IMC. Difficulty ramps from arithmetic progressions to interleaved, multi-rule, and second-difference patterns.
What it filters for. Structured hypothesis testing at speed: generate a candidate rule, check it against every given term, commit or discard. Candidates who "see" patterns but never verify bleed points on the ramp.
How to train. Learn the pattern families cold, then drill them shuffled and timed. Our IMC sequences guide covers the family taxonomy, and the timed sequences test runs the format with fresh questions each attempt.
Family 3: probability and quantitative quizzes
The format. Ten to forty questions of real probability, expected value, combinatorics, and logic, with more generous per-question time than the speed families. This is the style associated with probability-heavy firms' screens and with quant research roles broadly: multiple choice or numeric entry, and the questions have actual mathematical content.
What it filters for. The same core the later interviews test: conditional probability without traps sprung, expected value decompositions, quick combinatorial counting. It is the most syllabus-like family, which makes it the most preparable.
How to train. Work the curriculum in difficulty order: probability, expected value, combinatorics, then mixed timed sets. The free tier's easy questions are the on-ramp; the full bank's harder tiers match what the stronger screens actually ask. Our prep roadmap sequences the whole thing.
Family 4: game-based and behavioral assessments
The format. Interactive exercises rather than question lists: short games probing reaction time, working memory, risk decisions with real trade-offs, plus personality and work-style questionnaires. Several large firms, and some third-party platforms firms license, use these either standalone or bundled with a numerical stage.
What it filters for. Traits that are hard to cram: processing speed, consistency, risk posture, honesty (the questionnaires embed consistency checks). That is precisely why the cramming advice is thin here.
How to train. You mostly do not, and pretending otherwise wastes prep hours better spent on families 1 to 3. What you can do: be rested, take it in one quiet sitting, answer personality items consistently and truthfully, and practice general speed-under-pressure through the games, which build the same fast-decision muscle without pretending to be the test.
Family 5: coding screens
The format. For quant developer and many quant research tracks: algorithm and data-structure problems on a coding platform, usually one to three problems in 60 to 120 minutes. Trading-role candidates sometimes meet a light version.
What it filters for. Standard software interview material plus, at trading firms, an added premium on correctness under time, edge cases, and complexity awareness.
How to train. Standard competitive-programming prep transfers directly; this is the one family where the general software interview ecosystem fully applies. If your target is trading rather than dev, confirm from the posting whether a coding stage exists at all before spending weeks here.
Which format will you face?
Three signals, in order of reliability:
- The invite email. It names the test, the time limit, and the rules. Read it completely before clicking start; the details (attempt limits, webcam proctoring, allowed materials) vary and matter.
- The role. Trading roles skew to families 1, 2, and 4; research roles to 3 and 5; developer roles to 5.
- The firm's reported history. Our firm guides summarize the interview and assessment style candidates report at each major firm, which is a good prior even when cycles shift details.
Universal rules that save real points
- One attempt. Assume every assessment allows exactly one, because most do. Practice runs happen on practice platforms, never on the real link.
- Environment. Desktop over laptop trackpad, wired internet if possible, silence, phone away, and the same hour of day as your strongest practice runs when scheduling is yours to choose.
- The clock starts before the test. Many platforms time out invitation links after days. Opening the email late has ended candidacies.
- Honesty. Assessments increasingly note monitoring and consistency checks. Outsourcing or tooling an answer is both a values problem and, practically, a trap: the next stage retests the same skill live.
A two-week base plan across formats
Weeks are scarce in a rolling-admissions season, so this plan builds the shared base first and specializes on demand:
- Days 1 to 3: locate yourself. One 80 in 8 benchmark run, one sequences test run, one easy-tier probability set. Note the weakest of the three.
- Days 4 to 10: daily 15-minute arithmetic drills (this base pays off in every family), plus your weakest family's dedicated guide and drills.
- Days 11 to 14: full-format timed runs on alternating days, environment locked to test conditions, and review of every wrong answer by category rather than in isolation.
When an invite lands mid-plan, collapse to that family's guide and its timed format, and see the 2026-27 timeline for how assessments slot into the wider season.
FAQ
Are trading firm assessments proctored?
Increasingly, candidates report webcam or screen monitoring on some firms' assessments, and consistency checks inside behavioral tests are standard. Assume you are observed and act accordingly; the invite states the specifics.
Can I use a calculator or paper?
For speed arithmetic, no calculator by every consistent report, and often no paper. Probability quizzes vary. The invite email is definitive; when unstated, prepare for the stricter case.
If I fail one firm's assessment, is my cycle over?
No. Results are per firm, and formats repeat across firms, so a miss at one becomes format-specific training data for the next. Within the same firm, candidates generally report one attempt per cycle.
Do firms tell you your score?
Usually not. Most candidates receive only advance-or-reject. This is why practicing on a platform that does show scores and percentiles, like the free benchmark, is the only reliable way to know your level before it counts.