Bus Delays vs Expected Arrival

This expected value question is about understanding how random arrival times interact with a schedule that sometimes gets delayed, and how that changes the distribution you actually experience. Even though the bus follows a simple timetable and the delays follow simple probabilities, the effective timeline seen by a random passenger is skewed in a subtle and unintuitive way. It is a classic twist used in quant prep to check whether candidates see beyond the naive averaging of scenarios.

It trains your grasp of continuous-time probability, stationarity, and length-biased effects, as well as your ability to translate a verbal description into the correct random variables. You must recognize that the viewpoint of someone arriving at a random time is different from the viewpoint of counting events. This is core quant interviews material, strengthening comfort with rigorous conditioning and expectation.

This matters for a quant interview because many models in trading, market microstructure, and risk involve random arrivals in time: orders, quotes, defaults, or price jumps. Misjudging how often you "see" certain states leads to biased estimates, wrong risk assessments, and poor pricing. Being able to reason precisely about these inspection effects shows real depth in probability, beyond formula plugging, which is exactly what high-end quant interviews are trying to detect.

What it tests

When events or intervals have different durations but occur with fixed probabilities, the probability of being in a given interval at a random observation time is proportional to the length of that interval, not just its frequency. This is known as the inspection paradox or length-biased sampling. The intuition is that longer intervals occupy more of the total timeline, so a random arrival is more likely to fall within them, even if they are rarer. To compute expected waiting times or similar quantities, you must weight each scenario by the fraction of total time it occupies, not just by how often it starts. This principle holds because random sampling from a continuous timeline is more likely to land in longer intervals, so their impact is magnified compared to their mere count.