The Optiver 80 in 8 Test, Explained

Every major trading firm runs a numerical screen, but Optiver's is the one with a reputation. Candidates call it the 80 in 8: eighty arithmetic questions, eight minutes, and a scoring rule that makes guessing expensive. It filters out most of the applicant pool before a human ever reads a CV, and most people who fail it do not fail on arithmetic. They fail on strategy, because they never ran the numbers on their own answering behavior.

This guide covers the format as candidates publicly report it, the expected value math that should govern every skip decision, the pacing reality of a six second budget, and a three week plan that takes you from a cold first attempt to a competitive score.

What the 80 in 8 Actually Is

The format has been described consistently in candidate reports for years:

• 80 questions of pure arithmetic. No word problems, no logic puzzles.
• 8 minutes on a single running clock.
• Scoring: $+1$ for a correct answer, $-2$ for a wrong one, $0$ for a skip.
• No calculator.

It sits at the front of Optiver's trading funnel, typically before any human interview. If you want the full picture of what comes after it, the Optiver firm page maps the later rounds. Details vary by office, role, and recruiting year, so treat everything here as community-reported rather than official.

The content itself is arithmetic you learned by age twelve. The test is not measuring whether you can do it. It is measuring whether you can do it at speed, under fatigue, with the judgment to know when you are about to be wrong.

Why Negative Marking Changes Everything

Skipping scores zero, so zero is your benchmark. When you attempt a question with probability $p$ of being right, your expected score is

$\mathrm{EV} = p \cdot (+1) + (1 - p) \cdot (-2) = 3p - 2$

Set that to zero and solve: $3p - 2 = 0$, so $p = \tfrac{2}{3}$.

That single line is the strategy of the entire test. Attempt a question only when you are at least two thirds sure of your answer. Below that confidence, the skip button pays better.

Run a few values to feel the slope:

• At $p = 0.25$, a blind guess on a four option layout: $\mathrm{EV} = 3(0.25) - 2 = -1.25$
• At $p = 0.5$, a coin flip: $\mathrm{EV} = -0.5$
• At $p = \tfrac{2}{3}$: $\mathrm{EV} = 0$, exactly breakeven with skipping
• At $p = 0.8$: $\mathrm{EV} = +0.4$
• At $p = 0.9$: $\mathrm{EV} = +0.7$

On an open answer format a blind guess sits near $p = 0$ and costs you almost the full two points. Either way the conclusion holds: one wrong answer erases two right ones.

Here is the comparison that converts most people. Candidate A attempts 60 questions at 80 percent accuracy: 48 right, 12 wrong, for a net of $48 - 24 = 24$. Candidate B attempts only 40 questions at 95 percent accuracy: 38 right, 2 wrong, net $38 - 4 = 34$. B answered twenty fewer questions and won by ten points.

This is why trading firms love the format. It is a calibration test wearing a math test's clothes. Knowing what you know, and folding when you do not, is most of the job description.

The Question Mix

Candidate accounts describe four recurring families:

• Integer arithmetic. Addition, subtraction, multiplication, division, mostly two digit by one or two digit.
• Decimals. The same operations with a decimal point waiting to be misplaced.
• Fractions. Adding, comparing, simplifying, converting to decimals and back.
• Percentages. Percent of a number, percentage increase and decrease.

Almost everything resolves in one or two steps. There is no trick content and no algebra. The difficulty is entirely the clock, which deserves its own arithmetic.

The Pacing Math

Eight minutes is 480 seconds. Spread across 80 questions, that is six seconds per question, including reading it, computing, and entering the answer.

Six seconds rules out working anything out in the school sense. There is no room for long division, carrying digits on paper, or trying a second method when the first stalls. Every answer has to come from retrieval, meaning known tables and anchors, or from a single rehearsed transformation.

In practice strong candidates run a two tier clock. Easy questions get banked in three to four seconds, which buys ten to twelve seconds for the genuinely awkward ones. The discipline is the cap: if a question has not resolved in about ten seconds, skip and move with zero emotion. A skipped question costs nothing. A stubborn one costs the three or four questions you never reached.

What Scores Actually Look Like

Optiver publishes no cutoffs, so everything in this section is community folklore and you should weight it accordingly.

The reported pattern is consistent, though. Cold first attempts from strong STEM students commonly land at a net score in the teens or twenties, and guess-happy first attempts sometimes come back negative. After a few weeks of structured practice, candidates report attempt counts climbing toward most of the paper with only a handful of misses. The people who advance tend to describe high attempt accuracy rather than heroic coverage.

Reported pass bars vary by office, role, and season, and forum numbers contradict each other, so do not anchor on any single figure. Anchor on process: every attempt above the two thirds threshold, no time sinks, and a net score that climbs week over week.

A Three Week Training Plan

Week 1: Accuracy before speed

Untimed or lightly timed drills, twenty to thirty minutes daily. The goal is to make retrieval instant:

• Multiplication tables cold to 12 by 12, stretching toward 20 by 20
• Squares through $25^2$
• Fraction and decimal anchors: halves, quarters, eighths, twelfths, in both directions
• Complements to 100 and 1000 for fast subtraction

Work mixed sets from the question bank and log every miss. If you would rather have the drills sequenced day by day, the prep tracks lay them out for you.

Week 2: The six second cadence

Introduce the clock in bursts: twenty questions in two minutes, several times a day. Two skills get built here. The first is executing techniques at speed. The second is skipping, which needs deliberate practice because it feels like failure. Run sets where you force a skip on anything unsolved after ten seconds, then review whether each skip was the right call. Keep the error log going; most people discover that one family, usually decimals or fractions, produces the bulk of their misses.

Week 3: Full simulations

Daily full length runs under exam scoring at the mental math simulator. Track two numbers separately: net score and attempt accuracy. If attempt accuracy sits below roughly 90 percent, you are attempting too much, so tighten your threshold. If accuracy is near perfect but the net is low, you are skipping questions you actually know, so loosen it slightly. Ease off the day before the real sitting. Sharp beats exhausted.

Five Techniques, Worked Digit by Digit

These come from the full mental math technique set. The five below repay practice time fastest on this specific test.

1. Multiply by 25 using quarters

25 is 100 divided by 4, so multiplying by 25 means appending two zeros and quartering.

48 times 25: append zeros to get 4800. Halve it: 2400. Halve again: 1200.

76 times 25: 7600, halve to 3800, halve to 1900.

The same shape covers 50 (append zeros, halve once) and 125 (append three zeros, divide by 8).

2. Difference of squares

When two factors sit symmetrically around a round center, use $(a - b)(a + b) = a^2 - b^2$.

47 times 53: both sit 3 away from 50, so the product is $50^2 - 3^2 = 2500 - 9 = 2491$.

62 times 58: both sit 2 away from 60, so $3600 - 4 = 3596$.

Train the spot itself: any pair with an even sum has a clean center halfway between them.

3. Fraction and decimal anchors

Memorize the eighths cold: $\tfrac{1}{8} = 0.125$, $\tfrac{3}{8} = 0.375$, $\tfrac{5}{8} = 0.625$, $\tfrac{7}{8} = 0.875$.

37.5 percent of 64: recognize 37.5 percent as $\tfrac{3}{8}$. Then 64 divided by 8 is 8, and 8 times 3 is 24. Two retrievals, no decimal arithmetic touched.

21 divided by 8: peel off 16, which gives 2, leaving remainder 5, and $\tfrac{5}{8}$ is 0.625. Answer 2.625 with no long division.

4. Squares ending in 5

For any number ending in 5, multiply the leading digits by one more than themselves, then append 25.

$85^2$: take 8, multiply by 9 to get 72, append 25: 7225.

$65^2$: 6 times 7 is 42, append 25: 4225.

It works because $(10n + 5)^2 = 100n(n + 1) + 25$.

5. Flip the percentage

Percent-of commutes: $a$ percent of $b$ equals $b$ percent of $a$, because both equal $\tfrac{ab}{100}$.

16 percent of 25: flip to 25 percent of 16, a quarter of 16, which is 4.

8 percent of 75: flip to 75 percent of 8, three quarters of 8, which is 6.

One flip routinely turns an ugly computation into a quarter or a tenth you already know.

Common Failure Patterns

• Guessing below breakeven. Attempts at coin flip confidence bleed half a point each; twenty of them costs an expected ten points.
• The 25 second question. Sunk cost in miniature. The clock does not refund stubbornness. Cap at ten seconds and skip.
• Decimal point slips. Speed produces magnitude errors more than digit errors. Saying the rough size first, about 20 rather than about 2, catches most of them.
• Misread operators. Under fatigue, minus signs read as plus signs. Slow your eyes on the operator even when everything else rushes.
• The minute six cliff. Candidates who only ever trained short bursts report accuracy collapsing in the final minutes. Full length simulation is the only fix.
• Switching methods mid-question. Pick the route in the first second and commit. If no route appears, that is exactly what skipping is for.

FAQ

Is the 80 in 8 multiple choice or typed answers?

Candidate reports include both, varying by region and recruiting cycle. Strategically it barely matters: even with four options, a blind guess has an expected value of $-1.25$, so the skip discipline stays the same. Train for typed answers and a multiple choice version becomes a comfort.

What score do you need to pass?

There is no official number, and community reports vary too much by office and season to anchor on. The candidates who advance generally describe attempting a large share of the paper with very few wrong answers. Build attempt accuracy first and coverage second, and the net score follows.

Can you use a calculator or scratch paper?

No calculator, as universally reported. Accounts on scratch paper differ, so train without it. At six seconds per question, writing is mostly a tax anyway; if a method needs paper, you need a different method.

Train the Way the Test Is Scored

Generic mental math apps will not teach you the skip. QuantPit's 80 in 8 simulator runs the full cadence with the plus one, minus two, skip zero scoring, regenerates fresh questions every run, and reports attempt accuracy next to your net score so you can tune your threshold instead of guessing at it. Pair it with the worked technique set, and once your net score stabilizes, stress test yourself across the rest of the timed screens before the real sitting.

QuantPit is independent of Optiver. Format details are as candidates publicly report them.