If we have a box full of coins, with one gold coin with probability 10% to get extracted.

Given this, is it possible to find out how many times in average I have to try extracting a coin until I get the gold coin?

Note: if I extract a coin and it's not the desired one I have to put it again in the box. And I can extract ONE coin at a time.

  • 2
    $\begingroup$ What are your thoughts, attempts? You should include these in every post. Also, phrases like "Probability problem!" in the title are not very informative, and it's not that exciting. Hehe. $\endgroup$ – Em. Jun 6 '16 at 21:02
  • $\begingroup$ Also, since it has been answered, you are describing a geometric distribution on the integers greater than $0$. $\endgroup$ – Em. Jun 6 '16 at 21:20
  • $\begingroup$ @probablyme To answer your request, this isn't actually a homework, this question came to my mind while I was playing a game where to get a secret object I had only 10% of chance.. I know it's pretty strange, but I was wondering how many times I needed to try and If there was actually a "realistic" probabilistic average of tries! $\endgroup$ – Ergo Jun 29 '16 at 21:58

With probability $0.1$, you'll get the coin in $1$ try (one success).

With probability $0.9\cdot0.1$, you'll get the coin in $2$ tries (one failure, one success).

With probability $0.9\cdot0.9\cdot0.1$, you'll get the coin in $3$ tries (two failures, one success).


The expected number of tries is thus:

$$\sum_1^\infty 0.9^{n-1}\cdot0.1\cdot n = 10$$

(In $n$ tries we have failed $n-1$ times ($0.9^{n-1}$) and succeeded once ($0.1$), and we need to sum over all possible values of $n$.)

This lines up with our intuition: if we flip a coin, we expect on average $2$ flips before we get tails, since the probability of tails of $\frac 12$. Here, since the probability of the gold coin is $\frac{1}{10}$, we intuit an expected waiting time of $10$.

Just to make this post self-contained, I'll show an evaluation of the sum here. Note that this is probably not the easiest way to evaluate the sum, so if anyone would like to edit this post with a more tangible way, feel free (this is just the first method that came to mind).

$$f(x) = \sum_1^\infty{x^n} = \frac{1}{1-x}, \ \ |x|<1$$ $$f'(x) = \sum_1^\infty{n x^{n-1}} = \frac{1}{(1-x)^2}, \ \ |x|<1$$

We want to evaluate:

$$0.1 \cdot \sum_1^\infty{n\cdot 0.9^{n-1}}$$

Plugging in to the second equation, we get:

$$0.1 \cdot \frac{1}{(1-0.9)^2} = 0.1 \cdot 100 = 10$$

  • $\begingroup$ Why are you saying that with probability 0.1 you'll get the coin in 1 try? Isn't that 10/100 so we have 10%, shouldn't it be 10 tries? $\endgroup$ – Ergo Jun 6 '16 at 21:10
  • $\begingroup$ Everything you said is correct, so I'm not sure where your confusion is. There is a 10% (=$0.1$) chance we'll get the gold coin on try number one. If we don't (90%), there is a 10% chance we'll get the gold coin on try number two. We need to sum up over all possible values - in theory it might take infinitely many tries to get the golden coin. When we sum up over all values, the sum converges to $10$, meaning on average it will take $10$ tries to get the gold coin. $\endgroup$ – anonymouse Jun 6 '16 at 21:12
  • $\begingroup$ Ahh okay. There is still something else I don't understand why do we have to multiply by n in the sum? And how do you evaluate the limit of the sum? $\endgroup$ – Ergo Jun 6 '16 at 21:17
  • $\begingroup$ To your first question, we looking for the expected number of tries. So in the case of $2$ tries, the probability that this happens is $0.9\cdot0.1 = 0.09$, but this case takes... well, it takes $2$ tries. So we need to multiply the $2$ in. Essentially what we're saying is that the expected number of tries is the sum of the amount of tries multiplied by their respective probabilities: $E(X) = \sum{X\cdot P(X)}$. To your second question, I'll edit my post with the evaluation of the sum. $\endgroup$ – anonymouse Jun 6 '16 at 21:23
  • $\begingroup$ I've edited the post with an evaluation of the sum. Let me know if you have any lingering questions. $\endgroup$ – anonymouse Jun 6 '16 at 21:35

Let's say the expected number of trials to get a gold coin is $E$. On the first try, we either get a gold coin, with probability $1/10$, or we don't, with probability $9/10$. If we don't get the gold coin on the first trial, then we are back where we started, except that the expected total number of trials is now $E+1$. So $$E = \frac{1}{10} \cdot 1 + \frac{9}{10} \cdot (E+1)$$ Solve for $E$.


Way too detailed for such a simple problem.

It is this easy: $P(gold coin)=\frac{1}{10}$. Hence, after $10$ tries we can expect a gold coin on average.

  • 1
    $\begingroup$ I agree that intuition leads us to this answer, but I don't know if this really explains what's going on behind the scenes. For someone really lacking the intuition, there's no reason to believe $10$ is the answer when the number of tries could theoretically be infinite. $\endgroup$ – anonymouse Jun 6 '16 at 21:15

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.