Casino Dice Stop Rule

This question describes a sequential casino game built around repeated dice rolls, where the player faces a trade-off between locking in accumulated gains and risking everything to continue. It is a stylized optimal stopping problem with a simple Markovian state variable: your current total winnings. At each step, you either walk away with that state or accept an additional random draw that might increase the total or wipe it out. This sort of setup is common in quant interviews for trading and strat roles, where candidates must reason about dynamic decisions under risk using clean mathematical structure rather than intuition alone.

To answer correctly, a candidate must translate the narrative into a recursive value function and apply optimal stopping ideas from stochastic processes. The problem leans on conditional expectation, risk-neutral valuation, and backward induction over an infinite or unbounded horizon. Interviewers watch for correct formulation of the dynamic programming equation, consistent handling of probabilities and states, and recognition of when and why a threshold policy is optimal. They also look for numerical and algebraic discipline, and for clear explanation of the stopping rule in economic terms.

What it tests

This problem class is governed by the principle of optimal stopping in the presence of risk and recursion. The key structure is that at each decision point, you compare the immediate, certain value of stopping with the expected value of continuing, which itself depends on the future states you might reach. The recursion arises because the expected value of continuing is built from the possible next outcomes, weighted by their probabilities, and the process repeats until a stopping criterion is met. The optimal stopping threshold is where the expected value of continuing no longer exceeds the value of stopping. This threshold is not arbitrary: it emerges from balancing the potential incremental gains against the risk of losing everything, and is found by solving a recursive equation that encodes the game's transition structure.