Two Dice Expectation Game

This quant interview question is about conditional expectation in a setting where the sample space changes after the first random outcome. The structure makes you track how the distribution of possible values evolves once one option is removed, and how that affects the total you ultimately care about. It's a classic twist on simple dice intuition that forces you to think in terms of conditional distributions rather than naive symmetry.

It trains conditional expectation, the law of total expectation, and comfort with dependent random variables in multi-step experiments. You must keep a clean separation between the first outcome and the updated probability space for the second, while still reasoning about the combined sum. It also sharpens algebraic clarity and pattern recognition in expectations.

This matters for quant interviews and serious quant prep because you constantly value payoffs whose distribution changes after intermediate events. Interviewers use variants of this to test whether you truly understand conditional expectation structures, not just plug formulas. Problems like this sit at the core of quant interviews, from trading strategy modeling to risk scenarios, and good quant prep should include many similar multi-stage expectation questions.

What it tests

This problem class is governed by the law of total expectation, which allows us to break down a complex expectation into a sum of conditional expectations over all possible initial outcomes. When an action alters the sample space for subsequent steps (such as removing a value from a set), the expected value of future steps must be recalculated based on the new, reduced set. The key is that the expectation for the second step depends on the outcome of the first, but when all possibilities are averaged, the symmetry or uniformity of the original distribution often reasserts itself. This is because the process of removing each possible value and averaging over all such cases restores the original mean, provided the process is symmetric and unbiased. The principle holds because expectation is linear and does not depend on the order or dependency structure, as long as all conditional probabilities are accounted for correctly.