Roll Until Six Even Sum

This conditional probability question sits at the heart of quant prep for interviews because it blends a simple random experiment with a nontrivial stopping rule. Instead of a fixed horizon, the experiment ends at a random time, and the property of interest is a derived feature of the whole path, not a single outcome. That mix of path dependence and randomness is exactly what appears in many quant interview questions.

It mainly trains your ability to track an invariant as a process evolves, focusing on how each outcome updates a reduced state rather than the full history. You practice modeling the random evolution with a tiny state space, identifying how each new observation changes that state, and encoding the effect of the stopping condition in rigorous probabilistic terms.

This matters for quant interviews because real trading strategies and risk systems often depend on path properties under random horizons. Being fluent with these Markov-style abstractions and conditional probabilities shows you can turn a messy, evolving situation into a clean, analyzable model. It signals to interviewers that you can cut through noise, spot the right state variables, and reason about stochastic processes efficiently under time pressure.

What it tests

Problems involving repeated random processes with a stopping condition often reduce to analyzing how the process evolves under a simple invariant, such as parity. The key is that certain operations (like adding an even number) do not change the invariant, while others (adding an odd number) do. This allows the problem to be modeled as a Markov process with a small state space (here: even or odd sum), where the transition probabilities depend only on the current state and the next random event. The stopping condition (first 6) ensures the process is finite, and the symmetry or structure of the transitions allows us to set up and solve a recurrence or system of equations. The principle holds because the process's evolution depends only on the current state and the next random increment, not the full history.