Video roulette math puzzle

I recently heard of a game called Netflix roulette where people pick a video at random and watch it regardless of the result.

It got me thinking about probability and I came up with the following problem.

Bob loves the TV show Law & Order. Each day he picks an episode at random and watches it. Given there are 456 episodes of the show, how many days will it take Bob to watch the entire series on average?

Bonus: Figure out a formula for a show that has n episodes.

I will post a complete solution on Thursday. If you figure out the answer, leave a comment on how long it took you to solve it.

Small cases

Consider smaller cases to get an idea.

If a series has just 1 episode, it will take 1 day to watch the entire series.

What about 2 episodes? On the first day, Bob will watch one of the episodes. How long will it take him to watch the remaining episode on average?

We can solve for the number of days N as as a sum of two conditional events. If he picks the episode he has not seen (with probability 0.5), then the conditional expectation is 1 day. If he instead picks the episode he has seen, then he essentially loses a day, and he is back to the starting point–so the expectation is N + 1.

In other words,

N = 1 * Pr(picks episode he has not seen) + (N + 1) * Pr(picks episode he has seen)

N = 1 * 0.5 + (N + 1) * 0.5 = 0.5 N + 1

N = 2

Note that it takes Bob 2 days on average to watch the unique episode that he picks with probability 1/2.

Thus, it takes Bob an average of 3 days (1 day for the first episode, 2 days for the second) to watch a series with two episodes.

Solution

We can think about the problem in terms of rolling a die. Each day Bob picks a new episode randomly is essentially like Bob rolling a die where each face represents an episode number.

The question is: how many times on average must a 6-sided die be rolled until all sides appear at least once?

The first roll can be any of the faces. On the second roll, there are 5 remaining unique faces out of 6. Using the logic above, we can conclude it will take an average of 1 / (5/6) = 6/5 rolls until one sees a different face.

We continue the logic to calculate the number of rolls until a new face. As there are 4 remaining out of 6, this will take 6/4 rolls on average. Continuing the logic, we can conclude the total number of rolls it will take on average to reveal every face at least once is:

1 + 6/5 + 6/4 + 6/3 + 6/2 + 6/1 = 147/10 = 14.7

In other words, for a series with 6 episodes, it will take Bob about 15 days to watch every episode.

For a series with n episodes, the similar series is:

For n = 456, this sum is roughly 3,056.

For very large n, the series sum is roughly n * ln(n)–though this approximation for 456 yields 2792 so it is a very rough approximation.



Share this post:

| More

Previous post:

Next post:



  • Shadewalker

    LaTeX Notation

    \sum_{n = 1}^{q} 1 + \frac{n – 1}{q}

    Where n is the number of unique episodes he has already seen and q the total episodes. To lazy to calc it ;)

    5 – 10 Minutes

  • Shadewalker

    erm…. Definiton mistake, sry.

    n = the n-th episode he’s watching (1 for the first, 2 for the second etc.)
    q = total number of episodes (in this case, 456)

  • Dan

    Set up a simulation with random numbers and ran 20 times. The average was 2961 with a std dev of 450.

    Another method was to calculate the number of days required to watch a new show. For example, it would take 1 day to watch the first original show, but over 200 days to watch the last show. Not sure on the correct formula here, but I got answers of 2391 and 2601.

  • Frank

    I get 3055.6 as the expected duration. ~5 min

  • Faisal

    Simulation with 10000 runs, I get 3055.02 with SD 585.782.

  • tgt

    Do we want an average? a 90% confidence? A 99% confidence? I don’t approve of this question. The only valid answer is “We can’t know for sure. It could take as little as 456 days, or never be completed.

  • http://www.mindyourdecisions.com/blog/ Presh Talwalkar

    Thanks tgt, you are right. I forgot to say “on average.”

  • Matt

    Probability of watching one episode follows a geometric distribution: f(x_i) = (1-p)^k-1*p, p = (1/n), E(x_i) = 1/p = 1/1/n = n

    sum of E(x_i) over n episodes in n*n = n^2.

    In our example above, 456*456= 207936 on average

  • Anthony

    I came up with the same as Shadewalker, except that my terms were the ceiling of the fractions, with no +1. LaTeX:

    \sum_{i=1}^{n}\left\lceil\frac{n}{i}\right\rceil

    Hope I didn’t screw up, haha. Have fun!
    Anthony

  • http://www.mindyourdecisions.com/blog/ Presh Talwalkar

    Thanks all for the responses–I have posted a solution above.





Previous post:

Next post:

Other posts you may enjoy reading: