I was asked a question during a job interview. I don't think I managed to solve it properly during the interview, so I would like someone to explain the answer. The question was as follow:

  • There is a good with value $S$ that is uniformly distributed on [0,1]
  • You are placing a bid $B$. If $B$ is larger than or equal to $S$, you receive the good
  • Immediately after your bid is placed it is determined whether you receive the good or not
  • Immediately after that the value of the good doubles to $2S$

How should you place the bid $B$ to maximise your expected return? (I.e. maximize $2S-B$)

  • $\begingroup$ What expected value are you trying to maximize? If you just want to win the auction then you would bid $2$. $\endgroup$ – John Douma Mar 29 at 21:28
  • $\begingroup$ @JohnDouma I tried to clarify the question. You want to maximize the return of the auction. $\endgroup$ – highviolet Mar 29 at 21:33
  • $\begingroup$ Assume the expected ualue is for you "gain", i.e. $0$ if you did not receive the goods, and $2S-B$ is you get it. I got expectation $0$ no matter how you place the bid $B$. Probably I did not understand the problem correctly. $\endgroup$ – Yu Ding Mar 29 at 21:34
  • $\begingroup$ $\newcommand{\I}{\mathbb{I}}\newcommand{\E}{\mathbb{E}}$If you bid an amount $b$ (using lower case to avoid confusion with random variables), then your return (which is a random variable as it depends on the random $S$) is $(2S-b)\I(S\le b)$. Hence your expected profit is $$f(b):= \E[(2S-b)\I(S\le b)],$$ where $\I$ is an indicator function and $S\sim\mathsf{Unif}[0,1]$. Can you compute this expectation as a function of $b$ (use integration)? $\endgroup$ – Minus One-Twelfth Mar 29 at 21:43
  • $\begingroup$ @YuDing: I agree: your expected return is $0$, no matter what you bid. Perhaps this was the answer the interviewer wanted? $\endgroup$ – TonyK Mar 29 at 21:43

Your profit is $2S-B$ when $S \lt B$ and $0$ otherwise. Integrating over $S$, we get $$\int_0^B (2S-B) dS=(S^2-BS)|_0^B=0$$ Bid whatever you like in the interval $[0,1]$ and your expected profit is the same, $0$.

This seems a surprising conclusion, so we should try some things to validate it. If you bid $0$ you will never get the object, so your profit is $0$. If you bid $1$ and the initial value is $0$, you lose $1$. If you bit $1$ and the initial value is $1$, you win $1$. It is linear in between, so your expectation at a bid of $1$ is $0$. The problem is scale invariant. If you bid $B$ you lose $B$ when the initial value is $0$ and win $B$ when you barely get the object, so your expected profit is $0$.


Now that we all know the expected gain is $0$ no matter what you bid, here's a simpler explanation "after the fact"... :)

Conditioned on you winning with bid $B$, the initial value of the goods is $Uniform(0,B)$, so the final value is $Uniform(0,2B)$, with an expectation of $B$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.