Suppose $n$ friends are trying to decide who will pay for everyone's dinner. They decide to choose the person ( the 'loser' ) by playing 'odd man out'. Each person is given a coin that has probability $p$ of falling heads. Each person tosses his coin. A person is said to be a loser if his toss outcome is different from that of other $n-1$ person's tosses.

Suppose the friends repeat the game until there is a loser. What is the probability that the game would be repeated exactly $k$ times.

I know how to find the probability that there is exactly one loser ( using binomial distributions). By the k'th toss there would be one H versus $n-1$ T or vice versa. However the tosses before the kth toss could be randomly distributed. It could be 2 H and $n-2$ T, or 8 H and $n-8$ T. I'm not able to find a solution for this since the possibilities seem so random.



  • Compute the probability that there is exactly one loser for a particular round, call it $m$.

  • The number of rounds required would follow Geometric distribution, Geo($m$).

  • $\begingroup$ ah, thanks, I miss that. $\endgroup$ – Siong Thye Goh Nov 22 '17 at 4:12
  • $\begingroup$ Here q = 1 - m, and Geo(m) would be (q)^k-1 x m. Am I right ?. ( m stands for the probability of getting an odd man out in one round ) $\endgroup$ – madhavU Nov 22 '17 at 4:12
  • $\begingroup$ yes, $q^{k-1}m$ is right. $\endgroup$ – Siong Thye Goh Nov 22 '17 at 4:15
  • $\begingroup$ Thanks a lot Siong Thye Goh!! $\endgroup$ – madhavU Nov 22 '17 at 4:16

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.