25 Quant Trading Brainteasers with Complete Solutions
The 25 classic brainteasers that appear most often in prop firm and quant hedge fund interviews — each with a complete solution and the reasoning pattern it tests.
25 Quant Trading Brainteasers with Complete Solutions
Brainteasers in quant interviews aren't about tricking you. They test whether you can reason carefully under pressure, recognize structure in novel problems, and communicate your thinking while you work. The interviewer watches the process, not just the answer.
Here are the 25 most commonly seen brainteasers, with complete solutions and notes on what each one actually tests.
Section 1: Probability Puzzles
1. The Coin and Two Envelopes
You're given two envelopes. One contains twice as much money as the other. You open one and see $100. Should you switch?
The naive argument says switch: the other envelope has either $50 or $200, with equal probability. EV = (0.5 × $50 + 0.5 × $200) = $125 > $100. Always switch!
But if you should always switch, you could switch again after switching — an infinite loop. The paradox reveals a flaw: we can't actually assign equal probability to "$50 or $200" without knowing the prior distribution of amounts.
The correct answer: Without a prior on the amounts, you cannot compute the EV of switching. If you do assume a prior (e.g., amounts are drawn from a specific distribution), the paradox resolves and sometimes you should switch, sometimes not.
Tests: recognizing when probability is ill-defined without a prior.
2. The Coin Flip Game
You and I each flip a fair coin repeatedly. You win if your coin shows heads before mine does. What's P(you win)?
P(you flip first and win) = 1/2. P(we both flip tails, then continue) = 1/4. P(you win from there) = same as original.
Let p = P(you win). p = 1/2 + (1/4) × p → p − p/4 = 1/2 → 3p/4 = 1/2 → p = 2/3.
Tests: recursive setup for stopping-time problems.
3. The Three Doors (Monty Hall)
A car is behind one of 3 doors. You pick door 1. The host opens door 3 (goat). Switch or stay?
Switch. P(car behind door 2 | host opened door 3) = 2/3.
Intuition: You picked a 1/3 chance door. The host's action concentrates the remaining 2/3 probability on the door he didn't open.
Tests: conditional probability; the classic answer surprises most people.
4. The Drunk Man
A man walks home. He takes one step forward or one step backward each minute, with equal probability. His home is 1 step away. What's the probability he ever reaches home?
This is the symmetric random walk starting at position 1, with an absorbing barrier at 0. P(reach 0 from 1) = 1 (the symmetric 1D random walk is recurrent — it returns to every position with probability 1).
But the expected time to reach home is infinite.
Tests: distinguishing recurrence probability from expected stopping time. The answer "1" surprises people who think "he might wander forever."
5. The Hat Problem
100 prisoners are each given a hat that is either red or blue (chosen uniformly and independently). They can see everyone else's hat but not their own. They must simultaneously guess their own hat color. What strategy maximizes the number of correct guesses?
The naïve approach gets 50 correct (everyone guesses randomly). Can you guarantee more?
Strategy: Number the prisoners 0–99. Each person counts the hats they can see. Person k guesses: (k − count of red hats seen) mod 2. This guarantees exactly 99 or 100 correct guesses — all but one.
Why? If the total number of red hats is even, prisoner 0's formula is correct; everyone else also computes correctly based on the same parity. One prisoner might be wrong but exactly 99 are right.
Tests: cooperative strategies and parity arguments.
Section 2: Expected Value Puzzles
6. The Dice Game
You roll a fair die. If you don't like the result, you may roll once more (you must take the second roll). What's your optimal strategy and expected payoff?
Reroll if the first roll is less than E[second roll] = 3.5. Roll ≤ 3: reroll. Roll ≥ 4: keep.
E[payoff] = P(first roll ≤ 3) × 3.5 + P(first roll ≥ 4) × E[first roll | ≥ 4] = 0.5 × 3.5 + 0.5 × (4+5+6)/3 = 1.75 + 2.5 = $4.25
Tests: dynamic programming setup, optimal stopping.
7. The Coin and Jar
You repeatedly flip a fair coin. If heads, you win $1. If tails, you lose $1. You start with $3 and play until you either reach $5 or go broke. What's P(reaching $5)?
Gambler's ruin formula: P(reach target N from position k) = k/N (for symmetric random walk).
P(reach $5 from $3) = 3/5 = 60%.
Tests: gambler's ruin; the clean formula surprises people who try to enumerate.
8. The Secretary Problem
You interview n candidates in random order. After each interview, you must immediately accept or reject. You want to maximize P(hiring the best candidate). What's the optimal strategy?
The optimal strategy: reject the first n/e candidates (≈ 37%), then hire the next candidate who is better than all you've seen.
P(best hire) → 1/e ≈ 37% as n → ∞.
For n = 10: reject first 3–4, then hire the next best-so-far. P ≈ 40%.
Tests: optimal stopping theory; the 1/e answer is elegant and memorable.
9. The Unfair Coin
You have a biased coin (unknown probability p of heads). How do you simulate a fair coin flip?
The Von Neumann method: flip twice. If HT → output heads. If TH → output tails. If HH or TT → repeat.
P(HT) = p(1−p). P(TH) = (1−p)p. These are equal → each conditional on getting a result, equal probability of heads or tails.
Expected flips per output: 1/(2p(1−p)). Worst case (p = 0.5, fair coin): 2 flips per result. Best case: near 0 or 1, very inefficient.
Tests: symmetry arguments; this appears in probability and algorithms interviews.
10. The Urn Switching Problem
Urn A has 2 white, 1 black ball. Urn B has 1 white, 2 black. You pick an urn uniformly at random and draw a white ball. P(you drew from urn A)?
P(A) = P(B) = 1/2. P(white | A) = 2/3. P(white | B) = 1/3.
P(white) = (2/3)(1/2) + (1/3)(1/2) = 1/3 + 1/6 = 1/2.
P(A | white) = (2/3 × 1/2) / (1/2) = (1/3)/(1/2) = 2/3.
Tests: Bayes' theorem; drawing white makes urn A more likely.
Section 3: Logic and Reasoning Puzzles
11. The Bridge and Torch
4 people must cross a bridge at night. They have one torch. At most 2 can cross at a time. Crossing times: A=1 min, B=2 min, C=5 min, D=10 min. What's the minimum time for all to cross?
Naive: send A back each time. Cost: 1+2+1+5+1+10 = 20 min.
Optimal: the slow pair (C and D) should cross together.
- A and B cross: 2 min. (Total: 2)
- A returns: 1 min. (Total: 3)
- C and D cross: 10 min. (Total: 13)
- B returns: 2 min. (Total: 15)
- A and B cross: 2 min. (Total: 17)
Minimum: 17 minutes.
Tests: recognizing the optimal structure: always have the fastest person escort the torch back.
12. The Water Jugs
You have a 5-liter and a 3-liter jug. No markings. How do you measure exactly 4 liters?
Fill 5L. Pour into 3L (fills it). 2L left in 5L. Empty 3L. Pour 2L into 3L. Fill 5L again. Pour from 5L into 3L until full (takes 1L). 4L remains in the 5L jug.
Steps: 5→3→empty 3→pour→5→3. Total 6 transfers, 4 liters in the 5L jug.
Tests: state-space problem solving; systematic search.
13. The 12 Balls
You have 12 balls identical in appearance. One is either heavier or lighter. Using a balance scale, find the odd ball and determine if it's heavy or light in exactly 3 weighings.
This is a classic information theory problem. 3 weighings produce 3³ = 27 possible outcomes, enough to distinguish 24 possibilities (12 balls × heavier or lighter).
The strategy is complex to write out fully, but the logic is: each weighing should split the remaining possibilities into three equal groups.
First weighing: put 4 balls on each side, leave 4 off.
- If balanced: odd ball is in the 4 unweighed; next 2 weighings identify it.
- If one side heavy: compare 3 from the heavy side plus a known-good vs 3 from the light side plus a known-good.
Tests: information-theoretic reasoning; the 3-weighing proof is a classic.
14. The Defective Lightbulbs
You have 3 light switches in the basement controlling 3 bulbs upstairs. You can go upstairs exactly once. How do you identify which switch controls which bulb?
Turn switch 1 on for 5 minutes, then off. Turn switch 2 on. Go upstairs:
- Bulb is on → switch 2.
- Bulb is off but warm → switch 1.
- Bulb is off and cold → switch 3.
Tests: using physical properties (heat) beyond the stated constraints.
15. The Circular Firing Squad
N soldiers stand in a circle. Each soldier simultaneously shoots the soldier to their left. After the shooting, some might survive (if shot from both sides). For what values of N does exactly 1 soldier survive?
Soldiers survive if their left and right neighbors both shoot someone else (i.e., neither the soldier to their left-of-left nor their right-of-right targets them).
Actually: each soldier shoots left. Soldier k is shot by soldier k+1. Every soldier is shot. 0 soldiers survive for all N ≥ 1 (every soldier is shot by the person to their right).
Wait — if they simultaneously shoot and are shot, they all die simultaneously. So 0 survivors for all N.
The interesting version: they shoot the nearest armed soldier. This is more complex and leads to the "Josephus problem."
Tests: careful reading of the problem statement and distinguishing "shoot" from "survive."
Section 4: Market and Trading Puzzles
16. The Mispriced Coin
You're market-making on a coin flip. Someone walks up and buys at your ask price. What does this tell you?
If your market is efficient (correct bid/ask), the buyer may be: (1) uninformed and buying at fair value, or (2) informed and knowing the coin is biased toward heads.
The act of buying at your ask gives you information: it's more likely you're dealing with an informed buyer (or someone who misjudges probability) than a pure random decision. A rational uninformed buyer would be equally likely to buy or sell.
The practical lesson: Aggressive order flow (buyers hitting your ask, sellers hitting your bid) is adverse selection. Adjust your mid upward after seeing buyers consistently hit your ask.
Tests: market microstructure intuition and adverse selection.
17. The Arbitrage
A stock trades at $100 in New York and $101 in London. What do you do?
Buy in New York, sell in London. Lock in $1 profit per share with no risk.
In practice, the arbitrage disappears within milliseconds because: currency risk (FX rates), transaction costs, execution risk (prices may move while you execute), regulatory differences.
The real answer: you buy in NY and sell in London simultaneously, hedging FX risk. The strategy works only if $1 > transaction costs + FX hedging cost + execution risk.
Tests: basic arbitrage reasoning and awareness of practical frictions.
18. The Option Puzzle
A stock is at $100. A call option with strike $110 and a put option with strike $90 both have the same premium. Is this fair? What does it imply about the market?
For a standard lognormal stock, out-of-the-money puts (strike < current price) and calls (strike > current price) at equidistant strikes do NOT have the same premium in general. The put-call symmetry at equidistant strikes holds only for specific distributions.
If they're priced equally, it implies: (1) the market expects the same probability of the stock being at $90 vs $110 at expiration, OR (2) the market is pricing in negative skew (fearing downside more).
In practice, OTM puts on equities almost always cost more than equidistant OTM calls — the volatility skew. Equal pricing would be unusual and might indicate an arbitrage opportunity.
Tests: options intuition and understanding of volatility skew.
19. The Market Maker's Dilemma
You quote a bid of $99.90 and ask of $100.10 on a stock worth $100. A large institutional buyer comes in and hits your ask for 100,000 shares. What do you do next?
You're now short 100,000 shares you need to buy back. The institutional buyer likely has information (or is executing a large order for reasons that might move the price).
Immediately: (1) move your ask up (don't offer more at $100.10 into the same buyer), (2) your bid should also move up (the buyer's action is informative — price is more likely going up), (3) begin buying to cover your short at current prices before the price moves further.
The lesson: Large directional order flow changes your estimate of fair value. Update your market.
Tests: market-making thinking and dynamic bid/ask adjustment.
20. The Convergence Trade
Two securities historically trade at a spread of 10 cents. Today the spread is 50 cents. You buy the cheap one and sell the expensive one, betting on convergence. What are the risks?
Risks:
- The spread widens before converging — you can be forced to unwind at a loss (this is what happened to Long-Term Capital Management).
- The spread never converges — the historical relationship breaks down (structural change in one of the securities).
- Funding risk — you might not be able to hold the position until convergence if you run out of capital.
- Correlation risk — if both securities fall simultaneously (instead of converging), you lose on both legs.
Tests: risk identification beyond just the expected value of the trade.
Section 5: Hard Puzzles
21. The 100 Prisoners Problem
100 prisoners, numbered 1–100. 100 boxes in a room, each containing a random number 1–100. Each prisoner may open 50 boxes. They must all find their own number. No communication after they start. What's the best strategy?
Naive strategy: Each opens 50 random boxes. P(each finds their number) = 1/2. P(all find it) = (1/2)^100 ≈ 10^.
Optimal strategy: Each prisoner starts at the box numbered with their own number, then follows the cycle: open the box whose number is shown until you find your own number or exhaust 50 opens.
P(success) ≈ 31% (much better than 10^). The key: the strategy fails only if the permutation has a cycle longer than 50 steps, which occurs with probability ~31% for random permutations.
Tests: counterintuitive probability and cooperative strategy.
22. The Pirates and Gold
5 pirates rank themselves 1 (most powerful) to 5. They divide 100 gold coins. The most powerful pirate proposes a split; if ≥50% vote yes, it passes. Otherwise, he's thrown overboard. What does pirate 1 propose?
Working backward:
- With 2 pirates: pirate 2 votes yes to anything (if 1 dies, pirate 2 has full power and gives himself all 100). Pirate 1 offers pirate 2 nothing and keeps 100.
- Actually: with 2 pirates, pirate 1 proposes 100/0; pirate 2 votes no and pirate 1 dies. So pirate 2 wins.
- With 3 pirates: pirate 3 must get pirate 4 or 5's vote. Pirate 4 gets 0 if we go to 2 pirates (he dies), so pirate 3 offers pirate 4 just 1 coin. Pirate 3 keeps 99.
- With 4: pirate 4 needs 2 votes (5 or 6 would vote yes for 1 coin): offers pirate 5 (who gets 0 in 3-pirate scenario) 1 coin, keeps 99.
- With 5: pirate 5 needs 3 votes (including himself). Offers 1 to pirates who get 0 in 4-pirate game (pirates 2 and 4). Keep 98.
Pirate 1 proposes: 98 coins for himself, 0 for pirate 2, 1 for pirate 3, 0 for pirate 4, 1 for pirate 5.
Pirates 1, 3, 5 vote yes → passes.
Tests: backward induction and game theory reasoning.
23. The Ant on a Rubber Band
An ant starts at one end of a 1-meter rubber band. It walks at 1 cm/s. After each second, the rubber band is uniformly stretched by 1 meter. Does the ant ever reach the other end?
Yes — the ant always reaches the other end (eventually). The key: the ant's relative progress (fraction of band covered) increases each second by (1 cm) / (current length).
After n seconds: progress = Σ_^ [1/(100 + 100k)] = (1/100) × Σ 1/(1+k) = (1/100) × H_n (harmonic numbers).
Since H_n → ∞, the progress eventually exceeds 1 = 100%. The ant makes it.
Tests: harmonic series divergence; the counterintuitive result requires knowing that the series diverges.
24. The Two Generals Problem
Two armies must attack simultaneously or not at all. They can only communicate via messengers who might be captured. Can they guarantee coordination?
No — it's provably impossible with unreliable communication. No matter how many confirmation messages are sent, the last message is unacknowledged, leaving uncertainty.
This is the "Two Generals' Problem" in computer science, and it proves that guaranteed consensus is impossible over an unreliable channel. TCP uses probabilistic approaches (timeouts, retransmissions) rather than guaranteeing delivery.
Tests: distinguishing "very likely" from "provably guaranteed" — important in risk management.
25. The Sleeping Beauty Problem
A fair coin is flipped. Heads: Sleeping Beauty is awakened Monday and the experiment ends. Tails: she is awakened Monday and Tuesday (with memory wiped between). When she wakes up, what probability should she assign to the coin being heads?
This is genuinely contested among philosophers and mathematicians:
"Thirder" view: P(heads | awake) = 1/3. There are 3 equally likely awakenings (Heads/Monday, Tails/Monday, Tails/Tuesday). Only 1 of 3 is heads. P = 1/3.
"Halfer" view: P(heads) = 1/2. The coin is fair; waking up provides no information about the outcome. P = 1/2.
The "halfer" view is philosophically defensible; the "thirder" view is more useful for betting. In an interview, name both views and your reasoning.
Tests: knowing that some problems are genuinely ambiguous, and the ability to hold two valid perspectives.
Common Patterns Across All 25
Work backward: Puzzles 6, 7, 8, 22 all yield to backward induction.
Symmetry arguments: Puzzles 3, 9, 10 have symmetry that simplifies computation.
Recursive structure: Puzzles 1, 2, 14 set up equations involving the unknown itself.
Information theory: Puzzles 5, 13 require counting states to know what's distinguishable.
Adversarial thinking: Puzzles 16, 19, 20 require thinking about what the other party knows.
Practice these patterns daily at Fermiq — the estimation and probability drills build the reflexes that make brainteaser approaches automatic in interview conditions.