Optiver "Intervals" — the exact 2026/27 test, free
The most distinctive module in the current Optiver OA: instead of a single number you submit an interval [lower bound, upper bound]. If the true answer falls inside (inclusive), your score for the question is lower ÷ upper — so tight, well-placed intervals pay the most and a miss pays zero.
Free, in-browser, no signup — opens the full assessment on its own page, exactly like the real screen.
What the real test looks like
The 18 reported questions mix six families: probability brainteasers on a 0–100 scale (dice duels, broken-stick triangles, a Buffon-needle trap whose answer is exactly 0), visual counting under time pressure (a 10×10 shape grid, a table of 42 coins, a 10×10 dice grid), visual measurement (a path with a scale bar), statistics on a 200-value score grid, time-series extrapolation (some deliberately noisy — the verb is “estimate”), and expected-value computations up to a weighted coupon collector.
Scoring & strategy
The scoring rule creates a real calibration exercise: [30, 60] around a true 42 scores 50%, [40, 50] scores 80%, and [43, 50] scores zero. Two structural consequences worth internalizing before the real thing: an interval with lower bound 0 scores 0 even when correct, and the rule rewards intervals positioned high, not just narrow. When you are certain, a zero-width interval [x, x] scores 100%.
Rebuilt from 2026/27 candidate-reported assessments for practice. QuantVault is not affiliated with Optiver; formats rotate between cycles.
FAQ
What is the Optiver Intervals test?
A module of Optiver’s online assessment where every answer is an interval [lower, upper]. If the true answer lies inside inclusive of endpoints, the score is lower divided by upper; otherwise zero. 18 questions, 60 seconds each, auto-submit on timeout.
How should I choose interval width?
Match width to your uncertainty: exact-solve questions deserve near-zero width, while counting and estimation questions deserve the width of your plausible error. Since the score is lower/upper, an interval sitting higher scores more than an equally wide one sitting lower.
What kinds of questions appear?
Probability brainteasers on a 0–100 scale, visual counting grids (shapes, coins, dice), path-length measurement against a scale bar, percentile reading from a 200-score table, noisy time-series extrapolation and expected-value questions like a weighted coupon collector.
Is this practice test free?
Yes — the full 18-question run and the 2-question sample are free in the browser, no signup.