← Jane Street Interview Insights

Jane Street·Data Scientist·Technical Phone Screen·Senior

SeniorPrefer not to say
Jun 2026

Summary

Jane Street Data Scientist interview with a game theory brain teaser that sounds deceptively simple but turns into a full-on Nash equilibrium derivation under pressure. The interviewer keeps asking 'why' until you've either proved your answer or fallen apart.

Questions Asked (6)

Q1

Two players each secretly pick an integer from 1 to 100. The larger pick has 10 subtracted from it, then that result is compared to the smaller pick. Whichever value is larger wins. What is the optimal strategy, and can you prove it?

Algorithms & Data StructuresTechnical Trade-offs
Author's notes

This one wrecked me a little.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

First, reframe the game as a payoff matrix and identify dominated strategies. Then, look for a mixed-strategy Nash equilibrium, likely with support on a small set of numbers, and prove optimality by showing indifference and no profitable deviations.

Pro tip: In zero-sum games with a penalty for being too high, the equilibrium often involves a geometric distribution; test small cases to spot the pattern before proving it.

1. Define the game and payoffs

Clearly state the rules: each player picks an integer 1-100; the larger pick is reduced by 10, then compared to the smaller pick; the larger resulting value wins. Define the payoff as +1 for a win, -1 for a loss, 0 for a tie.

2. Identify dominated strategies

Show that picking 1 is weakly dominated (e.g., picking 2 does at least as well against all opponent choices). Similarly, argue that very high numbers are risky because subtracting 10 may make them lose to slightly smaller numbers.

3. Find the mixed-strategy equilibrium

Assume a mixed strategy with support on a set of numbers. Set up indifference conditions: the expected payoff of each pure strategy in the support must be equal. Solve for probabilities, often leading to a geometric distribution.

4. Prove optimality

Verify that no pure strategy outside the support yields a higher expected payoff against the proposed mixed strategy. Also, show that the opponent cannot exploit the strategy by deviating.

5. Generalize and discuss implications

Mention how the solution changes with different bounds or penalties, and relate to real-world applications like bidding or pricing strategies.

Key Points to Mention

  • Zero-sum game and mixed-strategy Nash equilibrium
  • Indifference principle: opponent's expected payoff equal across support
  • Geometric distribution as equilibrium strategy
  • Dominated strategies and iterated elimination
  • Proof by contradiction: no profitable deviation
  • Connection to auction theory or competitive bidding

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q2

For the number-picking game: when exactly does the player who chose the larger number win, lose, or draw? Work it out as a function of the gap between the two picks.

Algorithms & Data Structures
Author's notes

Actually the part I got cleanest.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

First, clarify the rules of the number-picking game, including the range of numbers and the winning condition. Then, derive the outcome as a function of the gap by considering the parity of the gap and the positions of the numbers relative to the midpoint. Finally, express the conditions for win, lose, or draw in terms of the gap and the smaller number.

Pro tip: Demonstrate structured thinking by breaking the problem into cases based on the parity of the gap and the position of the smaller number relative to the midpoint. This shows analytical rigor and the ability to handle edge cases, which is highly valued at Jane Street.

1. Clarify the game rules

Confirm the range of numbers (e.g., 1 to N), the objective (e.g., to have the larger number), and any constraints. This ensures you and the interviewer are aligned before diving into analysis.

2. Define variables and outcome conditions

Let the smaller pick be x and the larger be y, with gap d = y - x. Define the winning condition for the player who chose y in terms of x and d, considering the game's mechanics.

3. Analyze parity and midpoint effects

Examine how the parity of d and the position of x relative to the midpoint of the range affect the outcome. Use symmetry arguments to simplify the analysis.

4. Derive win/lose/draw conditions

Based on the analysis, express the exact conditions for the larger-number player to win, lose, or draw as a function of d and x. Consider edge cases where x is near the boundaries.

5. Validate with examples

Test the derived conditions with small numerical examples to ensure correctness and to illustrate the logic clearly.

Key Points to Mention

  • The importance of clarifying the game rules upfront, especially the range of numbers and the winning condition.
  • The role of parity of the gap in determining the outcome.
  • The significance of the midpoint of the range and how it affects the outcome based on the smaller number's position.
  • The use of symmetry to simplify the analysis (e.g., reflecting numbers around the midpoint).
  • Edge cases such as when the smaller number is at the boundary or when the gap is zero (impossible if numbers are distinct).
  • The need to express the final conditions succinctly as a function of the gap and the smaller number.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q3

Prove that no pure (deterministic) strategy is optimal in this game, and derive the lowest number a rational player should ever consider picking.

Algorithms & Data StructuresTechnical Trade-offsAdaptability & Ambiguity
Author's notes

The dominance argument for the lower bound is actually clean once you see it: any pick below some threshold is strictly dominated by shifting up by the penalty amount, because you never do worse against any opponent pick and sometimes do strictly better.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

First, clarify the game rules and the definition of optimality. Then, use a proof by contradiction: assume a pure strategy is optimal, and show that an opponent can exploit it by choosing a number that beats it. Finally, derive the lowest rational number by iterated elimination of dominated strategies, typically leading to a threshold like 0 or 1 depending on the game.

Pro tip: Connect the game to real-world strategic interactions, such as bidding in auctions or pricing decisions, to show practical relevance. Emphasize that mixed strategies are essential in competitive settings where predictability is a disadvantage.

1. Clarify the game and optimality

Restate the rules, including the number range, winning condition, and what 'optimal' means (e.g., maximizing expected payoff). Confirm assumptions like rationality and common knowledge.

2. Prove no pure strategy is optimal

Assume a pure strategy (a fixed number) is optimal. Show that if everyone else plays it, an opponent can choose a number that wins more often, contradicting optimality. Use a symmetry or best-response argument.

3. Derive the lowest rational number

Apply iterated elimination of dominated strategies: start from the highest possible number and reason downwards. Show that any number above a certain threshold is dominated, leading to the lowest rational choice.

4. Discuss mixed strategies and equilibrium

Explain that the optimal strategy is mixed, often uniform over a range. Mention Nash equilibrium and how it ensures no pure strategy can be exploited.

5. Summarize and connect to data science

Conclude with the lowest number and relate the reasoning to data science concepts like game theory, decision theory, and strategic thinking in modeling.

Key Points to Mention

  • Definition of pure vs. mixed strategies
  • Proof by contradiction: any pure strategy can be exploited
  • Iterated elimination of dominated strategies
  • Nash equilibrium and mixed strategy equilibrium
  • Rationality and common knowledge assumptions
  • Application to real-world competitive scenarios (e.g., auctions, pricing)

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q4

Among the numbers that survive the dominance elimination, is there one that beats all the others? Describe the win/loss structure within that range and find the optimal mixed strategy.

Algorithms & Data StructuresTechnical Trade-offs
Author's notes

This is where I basically ran out of steam.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

First, clarify the dominance elimination process and identify the surviving numbers. Then, analyze the win/loss structure by constructing a tournament matrix and checking for a Condorcet winner. If none exists, formulate the problem as a zero-sum game and solve for the optimal mixed strategy using linear programming or iterative methods.

Pro tip: In game theory interviews, always connect the solution to practical implications, such as market making or resource allocation, to show you understand the business context.

1. Clarify the elimination process

Ask clarifying questions to understand how dominance elimination works and what 'survive' means. Confirm the set of surviving numbers.

2. Construct the win/loss matrix

For each pair of surviving numbers, determine which one beats the other based on the given rules. Create a matrix where entry (i,j) is 1 if i beats j, -1 if j beats i, and 0 if tie.

3. Check for a dominant strategy

Look for a number that beats all others (a Condorcet winner). If found, that number is the pure optimal strategy. If not, proceed to mixed strategies.

4. Formulate as a zero-sum game

Set up the problem as a two-player zero-sum game where the payoff matrix is the win/loss matrix. The optimal mixed strategy is the Nash equilibrium of this game.

5. Solve for optimal mixed strategy

Use linear programming (e.g., simplex method) or iterative methods (e.g., fictitious play) to find the mixed strategy that maximizes the minimum expected payoff. Interpret the probabilities.

Key Points to Mention

  • Dominance elimination and its effect on the strategy space
  • Win/loss structure and tournament matrices
  • Condorcet winner and its absence
  • Zero-sum game formulation and minimax theorem
  • Linear programming for solving matrix games
  • Practical interpretation in trading or market making contexts

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q5

How does the optimal strategy change if the penalty subtracted from the larger pick is k instead of 10, or if the range goes up to N instead of 100?

Algorithms & Data StructuresTechnical Trade-offs
Author's notes

Follow-up I mostly handwaved.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

First, clarify the underlying game: two players alternately pick numbers from 1 to 100 without replacement, and the player who picks the larger number pays a penalty k (originally 10) to the other. Then analyze how the optimal strategy depends on k and N by considering the trade-off between picking a high number to win the round and the penalty cost. Finally, discuss how the strategy changes as k varies (e.g., for small k, picking high is more attractive; for large k, picking low to avoid penalty may be better) and as N increases (e.g., the game becomes more about relative ranks and the penalty becomes less significant relative to the range).

Pro tip: Emphasize that the optimal strategy is not just about maximizing the chance of picking the larger number, but about maximizing expected net payoff, which depends on both the probability of winning and the penalty size. Also, note that as N grows, the penalty k becomes relatively smaller, so the strategy converges to simply picking the highest available number.

1. Clarify the game and payoff structure

Restate the rules: two players alternately pick numbers from 1 to N without replacement; the player who picks the larger number pays a penalty k to the other. Confirm that the payoff for a round is +1 (or +k?) for the winner and -k for the loser, or similar.

2. Analyze the base case (k=10, N=100)

Explain the known optimal strategy for the original parameters: likely a threshold strategy where you pick the highest number if it's above some threshold, otherwise pick the lowest to minimize penalty.

3. Vary k and determine threshold shifts

Discuss how changing k affects the threshold: as k increases, the penalty for losing becomes larger, so you become more conservative and may pick lower numbers to avoid being the larger pick. As k decreases, you become more aggressive.

4. Vary N and consider scaling

Explain that as N increases, the relative impact of k diminishes, so the optimal strategy approaches simply picking the highest available number. Conversely, for small N, the penalty is more significant.

5. Summarize the general principle

Conclude that the optimal strategy balances the expected gain from winning against the expected penalty from losing, and that the balance depends on the ratio k/N or similar.

Key Points to Mention

  • The game is a zero-sum game with asymmetric payoffs due to the penalty.
  • The optimal strategy likely involves a threshold: pick the highest number if it exceeds some value, otherwise pick the lowest.
  • As k increases, the threshold increases (more conservative play) because the penalty for losing is larger.
  • As N increases, the penalty k becomes relatively less important, so the strategy converges to picking the highest number.
  • The ratio k/N or k/(N-1) may determine the optimal threshold.
  • Consider edge cases: k=0 (no penalty, just pick highest), k very large (avoid picking high at all costs).

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q6

If your opponent picks uniformly at random over 1 to 100, what is your best response and what is your expected payoff?

Algorithms & Data StructuresProduct Analytics & Metrics
Author's notes

Interesting pivot because now it's not adversarial, it's just optimization against a fixed distribution.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

First, clarify the game's payoff structure and rules, as the question is underspecified. Then, assuming a common setup like a symmetric zero-sum game (e.g., higher number wins), compute the expected payoff for each pure strategy against the uniform distribution and identify the best response. Finally, discuss mixed strategies if needed and verify the solution.

Pro tip: Show that you recognize the ambiguity and ask clarifying questions before diving into calculations. This demonstrates careful thinking and prevents solving the wrong problem.

1. Clarify the game

Ask about the payoff structure: Is it a zero-sum game? What happens in ties? Are players choosing numbers simultaneously? This ensures you understand the problem correctly.

2. Define the payoff function

Assume a common payoff: you win if your number is higher than your opponent's, lose if lower, and tie if equal. Write the expected payoff for choosing a number x against a uniform opponent.

3. Compute expected payoff for each pure strategy

For a given x, the probability of winning is (x-1)/100, losing is (100-x)/100, and tying is 1/100. If win payoff is +1, lose is -1, tie is 0, then expected payoff = (2x-101)/100.

4. Identify the best response

The expected payoff increases with x, so the best pure strategy is x=100, yielding expected payoff 99/100 = 0.99. If ties are resolved differently, adjust accordingly.

5. Consider mixed strategies and equilibrium

If the game allows mixed strategies, note that against a uniform opponent, any pure strategy is a best response if the opponent is fixed. But in equilibrium, the opponent would not play uniformly if it's not optimal.

Key Points to Mention

  • Clarify the payoff structure and rules before solving.
  • Compute expected value by considering win, lose, and tie probabilities.
  • Recognize that against a uniform distribution, the highest number often maximizes expected payoff in a 'higher wins' game.
  • Discuss the difference between best response to a fixed strategy and Nash equilibrium.
  • Mention that if the game is symmetric, the uniform strategy might be part of a mixed equilibrium under certain payoff structures.
  • Show awareness of edge cases (e.g., ties, payoff values).

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.