Probability Brainteaser Walkthroughs: 12 Classic Quant Interview Problems with Reasoning Patterns

Probability Brainteaser Walkthroughs: 12 Classic Quant Interview Problems with Reasoning Patterns

Probability brainteasers are the gating round at quant interview superdays. The questions look elementary but trip up most candidates because the intuitive answer is usually wrong. Strong candidates don’t memorize answers — they recognize patterns, set up the math correctly, and reason through to the answer in 3–5 minutes. This guide walks through 12 of the most-asked classics with the reasoning pattern, the trap most candidates fall into, and the right approach.

1. Two Children: At Least One Is a Boy

Problem: Mr. Smith has two children. At least one of them is a boy. What is the probability that both are boys?

Common wrong answer: 1/2.

Right answer: 1/3.

Reasoning: Without information, the four equally-likely outcomes (in birth order) are BB, BG, GB, GG. Conditioning on “at least one is a boy” eliminates GG. Three remaining outcomes (BB, BG, GB) are equally likely. Of those, only BB has both boys. P(BB | at least one B) = 1/3.

Pattern: Conditional probability with selection bias. The information “at least one is a boy” reveals more than “the elder is a boy.” If the question said “the elder child is a boy,” the answer would be 1/2 (BB or BG, both equally likely).

2. Tuesday Boy Variant

Problem: I have two children. One is a boy born on a Tuesday. What is the probability the other is a boy?

Common wrong answer: 1/3 (extending Q1).

Right answer: 13/27.

Reasoning: Each child has 14 equally-likely (gender, day) combinations: 7 boy-days + 7 girl-days. So 196 total combinations for two children. Of those, the ones with at least one boy-Tuesday: count = 14 + 14 − 1 = 27 (boy-Tuesday in first slot ∪ boy-Tuesday in second slot, minus double-counted both-Tuesday). Of those 27, both-boy outcomes: 7 + 7 − 1 = 13. So P = 13/27.

Pattern: The “extra information” (born on Tuesday) updates the conditional probability in surprising ways. Adding more specific information about the known child shifts probability toward both-boys.

3. Birthday Problem

Problem: How many people are needed in a room so that there’s at least 50% chance two share a birthday?

Right answer: 23.

Reasoning: Compute P(no shared birthday) for n people = 365 × 364 × … × (365−n+1) / 365^n. Set this < 0.5 and solve. n = 22 gives ~52.4% chance of no shared; n = 23 gives ~49.3%. So 23 people is the threshold.

Pattern: Always compute the complementary probability when the direct calculation is hard. P(at least one match) = 1 − P(no match).

4. Coin Flip Pattern Waiting Time: HT vs HH

Problem: You flip a fair coin until you see HT. What’s the expected number of flips? Now until you see HH?

Right answer: E[HT] = 4. E[HH] = 6.

Reasoning for HT: Use states. State 0 = no progress; State 1 = just saw H. From State 0: flip → 1/2 H (go to State 1) + 1/2 T (stay at State 0). From State 1: flip → 1/2 H (stay at State 1) + 1/2 T (done). Let E0, E1 be expected additional flips. E0 = 1 + (1/2)E1 + (1/2)E0 → E0 = 2 + E1. E1 = 1 + (1/2)E1 + (1/2)(0) → E1 = 2. So E0 = 4.

Reasoning for HH: States: 0 = no progress, 1 = just saw H. From State 1: flip H → done; flip T → restart at State 0. E1 = 1 + (1/2)(0) + (1/2)E0. E0 = 1 + (1/2)E1 + (1/2)E0 → E0 = 2 + E1. Solving: E1 = 4, E0 = 6.

Pattern: Markov chain with states based on partial progress. HH takes longer than HT because failing on the H restart returns to State 0 (no H seen recently); failing on HT returns to State 1 (still have H).

5. The Ant on a Clock

Problem: An ant starts at the 12 position on a clock face. Each minute it moves clockwise or counterclockwise (each with probability 1/2) by one hour. What is the expected number of steps until it returns to 12?

Right answer: 12 (for a clock with 12 positions).

Reasoning: For a random walk on a cycle of n nodes, the expected return time is n. Quick proof: by symmetry, expected hitting time from any node to any specific other node is the same. Mean return time = n × P(at any specific position in steady state) = n × (1/n) × n? Actually the cleaner derivation: for a random walk on cycle Z/nZ, expected return time to start = n. For n = 12, answer is 12.

Pattern: Random walks on graphs. The expected return time relates to the stationary distribution. For symmetric walks on regular graphs, formulas are clean.

6. Pirates Dividing Gold

Problem: 5 pirates have 100 gold coins. The most senior pirate proposes a division. All pirates vote (proposer included); ties pass. If the proposal fails, the proposer is thrown overboard and the next-most-senior proposes. Pirates are rational, prefer surviving, and prefer more gold. What does the most senior pirate propose?

Right answer: [98, 0, 1, 0, 1] (Pirate 1 keeps 98, gives 1 to Pirate 3 and Pirate 5).

Reasoning: Backward induction. With 1 pirate (P5 alone): P5 takes all 100. With 2 (P4, P5): P4 needs only own vote (tie passes), so P4 takes 100. With 3 (P3, P4, P5): P3 needs 2 votes including own. P5 will get 0 from P4 in next round, so P3 offers P5 1 coin → P3 keeps 99, P5 gets 1, P4 gets 0. With 4 (P2, P3, P4, P5): P2 needs 2 votes. P4 will get 0 from P3 in next round, so offering P4 1 buys P4’s vote. P2 keeps 99, P4 gets 1. With 5 (P1, P2, P3, P4, P5): P1 needs 3 votes. P3 would get 0 from P2 next round; P5 would get 1 from P2 next round. P1 offers P3 1 (better than 0) and P5 1 (matching, but breaking tie at P1’s level — actually P5 would need 2 to be better; depending on tie-breaking convention, the answer adjusts). Standard convention: P1 keeps 98, gives 1 to P3, gives 1 to P5; P2 and P4 get 0.

Pattern: Game theory backward induction. Solve the smallest case first; build up.

7. Fair Coin from Biased Coin

Problem: You have a coin with unknown probability p of heads. How do you generate a fair (50/50) random outcome?

Right answer (von Neumann’s algorithm): Flip the coin twice. If HT, return Heads. If TH, return Tails. If HH or TT, repeat.

Reasoning: P(HT) = p(1−p). P(TH) = (1−p)p. These are equal regardless of p. So the conditional probability that the result is “Heads” given we returned a result is 1/2.

Pattern: Use symmetry to construct unbiased outcomes from biased inputs. The expected number of flips is 2/(2p(1−p)) = 1/(p(1−p)).

8. The Secretary Problem

Problem: 100 candidates apply for a job. You interview them one at a time. After each interview you must accept or reject (no recall). You want to maximize the probability of hiring the best candidate. What’s the optimal strategy?

Right answer: Reject the first ~37 (n/e ≈ 36.8) candidates. Then accept the first one better than all the previous.

Reasoning: The optimal strategy is to use the first 1/e fraction (~37%) as a sample to learn the distribution, then pick the next one that exceeds all sampled. P(best) ≈ 1/e ≈ 36.8% under optimal strategy.

Pattern: Optimal stopping. The “1/e law” appears in many contexts. The intuition: too few samples means you don’t know the distribution; too many samples means you’ve passed up the best one.

9. The 100 Prisoners Light Bulb

Problem: 100 prisoners are in solitary cells. They can meet once to plan strategy. Then, one at a time (in random order, with possible repeats), prisoners are taken to a room with a light bulb. Each prisoner can toggle the bulb or leave it. At any point, a prisoner can declare “all 100 prisoners have visited this room.” If correct, all are freed; if wrong, all die. What strategy gives a positive probability of success?

Right strategy: Designate one prisoner as the “counter.” Other prisoners turn the light ON only the first time they enter the room while finding it OFF. The counter turns it OFF and increments their count. When counter reaches 99 toggles-off-by-them, they declare victory.

Reasoning: Each non-counter contributes exactly one ON-toggle. Counter sees this only when they’re in the room and it’s ON, so counter’s count = number of distinct non-counter prisoners who have visited. When count reaches 99, all 99 non-counters have visited; with the counter included, all 100 have.

Pattern: Information theory and signaling under constraint. Distinguishing “first visit” from “later visit” is the key insight.

10. The Three Doors / Monty Hall

Problem: Three doors. Behind one is a car; behind two are goats. You pick door 1. The host (who knows what’s behind each door) opens door 3, revealing a goat. The host offers you a chance to switch to door 2. Should you?

Right answer: Yes, switch. P(car behind door 2 after switching) = 2/3.

Reasoning: Initial probabilities: P(car at door 1) = 1/3, P(car at door 2 or 3) = 2/3. After host opens door 3 with a goat, the 2/3 probability concentrates entirely on door 2. So switching wins with probability 2/3.

Pattern: Conditional probability where the host’s action is informative. Key: the host always opens a door with a goat, so the action gives information.

11. Bertrand’s Box Paradox

Problem: Three boxes. Box A has 2 gold coins. Box B has 2 silver coins. Box C has 1 gold and 1 silver. You pick a random box and a random coin from it; the coin is gold. What’s the probability the other coin in the same box is gold?

Right answer: 2/3.

Reasoning: Three coins are gold: both in Box A and one in Box C. Each is equally likely to have been the drawn coin. So P(drawn from Box A) = 2/3 and P(drawn from Box C) = 1/3. If from Box A, other coin is gold (P = 1). If from Box C, other coin is silver (P = 0). So P(other is gold) = 2/3 × 1 + 1/3 × 0 = 2/3.

Pattern: Conditional probability where intuitive “split the boxes evenly” thinking fails. The condition “drew gold” makes Box A more likely than Box C.

12. The Drunkard’s Walk on a Number Line

Problem: A drunkard starts at position 0. Each second he moves right or left with equal probability by 1 step. What’s the probability he’ll eventually visit position +n for some specific n > 0?

Right answer: 1 (with probability 1, he visits every integer eventually).

Reasoning: Symmetric random walk on integers is recurrent — almost surely returns to 0 infinitely often, and almost surely visits every integer infinitely often.

Pattern: Recurrence vs transience of random walks. 1D and 2D symmetric random walks are recurrent; 3D+ are transient. Famous result: “a drunk man will find his way home, but a drunk bird may get lost forever” (Pólya).

The Meta-Pattern: How to Approach Any Brainteaser

1. Identify the type: conditional probability, expected value, optimal stopping, random walk, game theory, combinatorics. Most brainteasers fit a category.
2. Set up notation cleanly: define events / states explicitly. Don’t compute in your head — write out P(A), P(B|A), etc.
3. Compute the complement when direct is hard: P(at least one) = 1 − P(none).
4. Use indicator variables for sums: for “expected number of X events,” sum P(each event happens).
5. Backward induction for game theory: solve smallest case first; build up.
6. States and recurrence for waiting-time: draw the state diagram; write expected-time equations; solve.
7. Sanity-check the answer: does it make sense? If asking “expected return time on cycle of size n” and you get 12 for n=12, sanity passes.

Frequently Asked Questions

See also: Breaking Into Quant Finance • Jane Street Interview Guide • Mental Math Drills for Trading Interviews