If I have an 11 sided coin where p(1) up to p(10) has probability 0.01. And p(11) has probability 0.9.

How do I calculate after how many throws p(1) up to p(10) each occurred at least a 100 times.

Also how do I create a binominal distribution of this, to determine the minimum and maximum 95 or 99% amount?

I am seeking for a general formula to determine this, it is assumed that p(1)-p(n) of the unbiased side all have the same probability though.

  • 3
    $\begingroup$ This sounds like some generalization of the coupon collectors problem or the occupancy problem. Did you already look into those topics and are you looking for further information or are you starting completely blank without knowledge of those more simple problems? $\endgroup$ Commented Feb 23, 2021 at 17:58
  • $\begingroup$ What sort of formula are you looking for? An exact solution or a practical approximation? $\endgroup$ Commented Feb 23, 2021 at 17:59
  • $\begingroup$ What do zou mean by 'how many times should a coin be thrown'? The number of times is obviously at least 10 x 100 times, and how many exactly will be different each experiment. But I guess that you mean something else. $\endgroup$ Commented Feb 23, 2021 at 18:01
  • $\begingroup$ Why would it be a binomial distribution? $\endgroup$ Commented Feb 23, 2021 at 18:04
  • $\begingroup$ "I am seeking for a general formula to determine this, " for this generalization it would help to explain in what sense you are looking. What is the underlying problem/motivation? (There will need to be made simplifications which need to be made with the application in mind) $\endgroup$ Commented Feb 23, 2021 at 18:07

2 Answers 2


You could approximate the problem by each side being independent and then the probability for all ten sides being above 100 is the probability for one side being above 100 to the power ten.

Below is a code example that compares this approximation with a simulation.


d = 100 ### cutoff-level
m = 20  ### number of bins/dice-sides
n = 5000 ### number of trials for histogram

### fucntion that repeats sampling untill all states are above d
sim <- function() {
  states = c(rep(FALSE,m),TRUE)
  cases = rep(0,m+1)
  count = 0
  while (prod(states) == 0) {
    count = count + 1
    s = sample(c(1:(m+1)), 
               size = 1, 
               prob = c(rep(0.1/m,m),0.9))
    cases[s] = cases[s] + 1
    if (cases[s] == d) {
      states[s] = TRUE

### do the simulation n times
sims <- replicate(n,sim())

### plot and compute histogram
h <- hist(sims, freq = 1, main = "histogram of simulation with curve of approximation",
     breaks = seq(20000,30000,400))

### compute approximate counts distribution function
ns = 1:max(h$breaks)
psingle <- 1-pbinom(d-1,ns,p=0.1/m)
pmult <- psingle^m

### add to historgram
counts = sapply( 1:length(h$mids), FUN = function(x) {
points(h$mids,counts*n, pch = 21, col = 1, bg = 0)
  • $\begingroup$ To calculate the lower boundary, does it also work like calculating it for one side and doing it to the 10th power? If yes then it's solved. $\endgroup$
    – Thomas
    Commented Feb 23, 2021 at 18:59
  • $\begingroup$ @Thomas I have no proof but intuitively I would say that the different sides will be relatively a lot independent due to the additional side which takes 90% of the coin flips (I would prefer to speak about dice rolls when there are more than two sides). $\endgroup$ Commented Feb 23, 2021 at 19:28
  • 1
    $\begingroup$ I will see whether I can add a quick simulation that makes sense. $\endgroup$ Commented Feb 23, 2021 at 19:29
  • $\begingroup$ Thank you very much for this explanation. Also thank you for creating an entire simulation to prove it, it's very appreciated. Solved. $\endgroup$
    – Thomas
    Commented Feb 23, 2021 at 20:46
  • $\begingroup$ @Thomas I must admit that the simulation is more like a heuristic. I wouldn't say that it is proved. While I have this intuition that one can use this approximation by regarding the cases as independent, I also have a feeling that there might be some small correction factor that is not too difficult. $\endgroup$ Commented Feb 23, 2021 at 21:08

You have to throw it at least 1000 times to have a positive probability of 100 for each. The probability is always less than 1 no matter how many time you throw the die.
Suppose I throw the die $N \ge 1000$ times and let $X_i$ denote the number of times side $i$ is observed. $(X_1,...X_{11})$ has a multinomial distribution.

$$P[X_1 \ge 100, X_2 \ge 100, ...X_{10} \ge 100]$$ $$=\sum_{x_1=100}^{N-900}\sum_{x_2=100}^{N-800-x_1}\sum_{x_3=100}^{N-700-x_1-x_2}...\sum_{x_{10}=100}^{N-\sum_{j=1}^9x_j}P[X_1 =x_1, X_2 =x_2, ...X_{10} =x_{10}]$$ $$=\sum_{x_1=100}^{N-900}\sum_{x_2=100}^{N-800-x_1}\sum_{x_3=100}^{N-700-x_1-x_2}...\sum_{x_{10}=100}^{N-\sum_{j=1}^9x_j}\frac{N!}{x_1!x_2!...x_{10}!(N-\sum_{k=1}^{10}x_k)!}p_1^{x_1}p_2^{x_2}...p_{10}^{x_{10}}p_{11}^{N-\sum_{k=1}^{10}x_k}$$

You definitely do not want to try calculating that unless N is close to 1000.

For large $N$, $(X_1,...X_{10})$ is approximately multivariate normal with mean $(0.01 N, ..., 0.01 N)$ and variance matrix that has $0.01\times N\times(1-0.01)$ on the diagonal and $-0.01^2N$ off diagonal. see here

With $N=11600$, there is about 50% chance of having observed 100 of each face from 1-10.
R program:


For 95% chance, use $N=12900$
For 99% chance, use $N=13600$

  • 1
    $\begingroup$ This is a good idea. But I tested it and found (as expected) that the skewness of the marginal distributions makes this a poorer approximation than the Binomial solution offered by Sextus Empiricus. $\endgroup$
    – whuber
    Commented Feb 23, 2021 at 21:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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