Board Thread:Game Discussion/@comment-38614970-20190223152313/@comment-38174099-20190303163226

Do you use any particular programming language? In astrophysics we use Fortran, it's quite old but still very fast in computational terms. I used to make benchmark tests with other programs that allow you to make graphics too without using a second program (e.g. RStudio, MathLab), but I don't use these for years.

Actually is highly probable that you already know this. Anyway, as you said, you don't know which is the underlying distribution. To find it out you could use a Chi-Squared test putting on x-axis the time and on y-axis the frequencies (or number of ads) to have an idea of how your distribution is made (shape) and to see if the ads duration is really random. I suppose that you won't find out a Gaussian centered in 15 s, because (I think) there are e.g. more ads of 30 s than 5 s.

You made a great work, and you could still do some different things and tests based on months to have more accuracy and precision. But I suggest you to don't go for them. It's not worth it. Maxino 77 wrote: I'm just not sure and probably won't bother it because the data gathering process is really a pain in the a@#. Exactly. It's too much time consuming.