Broken Stick Lengths: Max, Min, Mid

This question looks at a classic broken-stick experiment, where a continuous object is cut at random and one studies the resulting segment lengths. Here the focus is not on feasibility events (like forming a triangle) but directly on the expected values of the longest, shortest, and middle-length pieces. Candidates must reason about a continuous sample space of all possible cuts, understand how that induces a joint distribution on three segment lengths constrained by a fixed total, and then relate each of the three order statistics to regions in that space.

Solving it leans heavily on geometric probability, order statistics for continuous random variables, and multivariable integration over constrained regions. The interviewer is looking for comfort with representing random partitions as points in a simplex, setting up correct inequality descriptions for max, min, and median events, and exploiting symmetry to simplify the work. They also watch how cleanly the candidate translates the intuitive picture into integrals, handles boundaries, and checks that the three expectations are consistent with the fixed total length.

What it tests

When breaking an object at random points and analyzing the resulting segments, the problem is governed by the geometry of the sample space and the order statistics of the resulting lengths. The key is to represent the random cuts as points in a constrained region (often a simplex or triangle), and then to translate questions about the maximum, minimum, or median segment into inequalities that define subregions of this space. The expected values of these order statistics are determined by integrating over these regions, often leveraging symmetry or conservation (such as the total length being fixed). This approach generalizes to any problem where a whole is randomly partitioned, and the properties of the resulting pieces are of interest. The pattern holds because the randomness of the cuts creates a uniform distribution over the feasible region, and the order statistics partition this region into geometrically interpretable subregions.