Mental Math for Quant Trading Interviews: Accuracy Drills and Error Diagnosis

Mental math for a quant trading interview is not a contest to memorize the largest multiplication table. The useful skill is to choose a short calculation, keep units and direction intact, and catch a bad answer before it drives a decision. This guide gives you an original practice set and an error log you can reuse. It is a cross-firm preparation exercise, not a claim about any firm's current test, timer, passing score, or permitted tools. Your invitation is the authority for those details.

The governing habit is simple: estimate a range, compute, then check the sign and magnitude. A fast wrong answer is less useful than a slightly slower answer you can defend.

What should you calculate first under pressure?

Before multiplying, say what the answer means. Is it dollars, shares, a probability, a price change in basis points, or an expected payoff per trade? Then estimate an upper and lower bound. For 19% of 248, the result must be a little below 20% of 250 = 50; 47.12 is plausible and 471.2 is not. The estimate can be spoken in two seconds and prevents a misplaced decimal from masquerading as fluency.

Choose a transformation that preserves the exact quantity. 19% of 248 = 20% of 248 - 1% of 248 = 49.6 - 2.48 = 47.12. This is more reliable than multiplying 248 × 0.19 digit by digit. A candidate who gives a transparent shortcut allows the interviewer to follow the calculation and correct an assumption. If a problem asks for an approximation, say when you rounded and how large the resulting error could be. Do not present an estimate as an exact result.

In a trading context, the final number often supports a decision rather than ending the question. For example, if a contract costs 42 and has a 45% chance of paying 100 and otherwise pays 0, its expected payout is 45; expected profit is 3 per contract before fees. A common mistake is calling 45 the profit. The arithmetic is easy; the unit and baseline make it meaningful. This is an invented exercise, not a quoted market or employer question.

Original drill: twelve calculations with a diagnosis key

Use three passes rather than one frantic sprint. On pass one, write only your answers and an approximate bound. On pass two, inspect your strategy without the key. On pass three, compare with the exact values below and assign each miss a cause. Start untimed; later use a timer as a personal measurement. There is no universal quant-interview time target implied by this exercise.

Tier	Prompt	Exact answer	Fast route and likely trap
A	15% of 240	36	10% = 24, 5% = 12; watch percent-as-decimal errors.
A	48 × 25	1,200	Quarter of 48 × 100; check the zero count.
A	1/8 of 376	47	Halve three times; do not round early.
A	7.5% of 800	60	5% + 2.5%; keep percent and basis points distinct.
B	103 × 97	9,991	(100+3)(100-3)=10,000-9; sign matters.
B	35% of 180	63	30% + 5%; a nearby 60 estimate checks it.
B	2/3 of 270	180	Divide first, then multiply.
B	1.25 × 96	120	Add one quarter of 96; not 25% of 1.25.
C	Price moves from 80 to 84; percentage increase?	5%	Divide the gain 4 by original 80, not final 84.
C	Buy at 99.50, sell at 100.25; gross gain per unit?	0.75	Subtract price levels; name gross versus net.
C	0.4 × 12 + 0.6 × (-5)	1.8	Compute weighted outcomes and preserve the negative sign.
C	Odds imply 1/4; your estimate is 30%; edge in percentage points?	5	30% - 25%, not a 5% relative return.

All answers can be checked independently: the first tier has exact integer outputs, 103 × 97 uses difference of squares, and the weighted-outcome example is 4.8 - 3 = 1.8. The final row deliberately asks for percentage points. A relative increase from 25% to 30% would be 20%, a different quantity. If you answered 20, your arithmetic was not the issue; you solved the wrong question.

For a second pass, change one number while preserving the method: try 15% of 260, 47 × 25, and 0.4 × 12 + 0.6 × (-8). Your goal is to transfer a strategy, not remember the key. For a third pass, explain one row aloud using only the expression, unit, shortcut, and check. Spoken reasoning reveals the sign and baseline errors that silent practice hides.

Diagnose the error before drilling again

A wrong answer deserves a label. Read error: you used 30 when the prompt said 13. Model error: you divided by the final price rather than the starting price. Arithmetic error: the right operation was chosen but an intermediate product was wrong. Sign or unit error: you computed an expected payout but called it profit, or confused basis points with percent. Pacing error: you abandoned a simple shortcut for long multiplication, rushed the final check, or froze on a difficult item. These causes need different remedies.

Keep a small log with four fields: prompt, your answer, correct answer, and cause. Add the earliest point of divergence, not merely the final digit. If 35% of 180 became 54, did you omit the five-percent term, miscompute nine, or read thirty percent? Three identical wrong outputs could call for three different interventions. Repeat one corrected sibling problem immediately, then another later; repeating the original alone can create recognition without transferable skill.

For an arithmetic miss, practice the exact operation family: complements to 100, quarters and eighths, or two-digit multiplication near a round number. For a model miss, write a sentence defining the denominator or payoff before calculating. For a unit miss, annotate the expression with %, percentage points, basis points, dollars, or contracts. For a pacing miss, use a two-pass approach: answer quick, high-confidence items first when the actual assessment permits navigation; otherwise practice a brief reset and move on according to its rules. Never assume the platform allows skipping or revisiting.

Track both accuracy and time, but do not optimize time at the cost of systematic errors. A useful personal metric is correct / attempted alongside median seconds per problem and cause counts. If speed improves while model errors rise, the intervention failed. Measure by category, too: percentages may be strong while conditional probability remains weak. The log should tell you what to practice next, not merely award a score.

Shortcuts worth understanding, not just memorizing

Percentages become easier when decomposed around 10%, 5%, 1%, and halves. 17% of 350 = 10% + 5% + 2% = 35 + 17.5 + 7 = 59.5. Another exact route is 20% of 350 - 3% of 350 = 70 - 10.5. Prefer the route with fewer fragile steps for you. For a rough bound, 17% is between 15% and 20%, so the answer is between 52.5 and 70. If the result falls outside, stop before using it.

Products near a round number have useful structure: (a+b)(a-b)=a²-b², as in 103 × 97. For 48 × 25, multiplying by 25 is dividing by four then multiplying by one hundred. For 99 × 68, compute 100 × 68 - 68 = 6,732. These are algebraic identities, not tricks tied to one firm. Always check scale: 99 × 68 should be close to 6,800, never 680.

Fractions and odds require clear definitions. A 1/4 probability is 25%; a 3:1 payoff ratio could mean several things unless you define whether it describes odds against, profit on a win, or total payout. Ask for the payoff convention. When converting basis points, 100 basis points is one percentage point. A 20-basis-point change in a percentage rate is 0.20 percentage points, not 20%. For prices, clarify whether the interviewer wants absolute change, percent change, or return including fees. This small verbal check is usually more valuable than a spectacular calculation on the wrong quantity.

Move from arithmetic to a trading answer

A strong response has four parts. First, restate the event and payoff in plain language. Second, compute an approximate fair value or edge. Third, show a short exact calculation if the numbers allow it. Fourth, state one reason the numerical answer might not be enough: transaction costs, correlated outcomes, inventory, uncertain probabilities, or a different objective such as chance of ruin. The challenge is not to pad every response with risk jargon. It is to make the numerical result relevant to the actual decision.

Consider a contract that pays 100 on a win, costs 42, and has your estimated win probability 45%. Fair expected payout is 45; expected profit is 3 before fees. If the round-trip cost were 4, the expected net profit would be -1. A candidate who stops at “positive edge” without asking about costs has missed the economic question. If the 45% is uncertain, present a threshold: break-even win probability before fees is 42%; with cost 4, it is 46%. Since 45% lies between, the cost assumption changes the decision. This threshold reasoning is a reusable interview asset.

The arithmetic does not establish that the probability estimate is good. Ask where it came from and what evidence could move it. If an interviewer changes the payoff or probability, adapt the expression rather than starting from scratch: expected net profit per unit is p × payout - price - cost for this simplified all-or-nothing contract. The explicit formula helps keep payout, purchase price, and transaction cost from being counted twice.

Sources and Further Reading

• Optiver US Campus Recruiting Q&A — official recruiting context; it does not define a universal format for this original drill.