Finally! Some actual data…

This is an interesting exercise. Let’s take the first part of the table and, assuming binomial distribution of the chance to get the stranger (we don’t care much for anything else here), make some crude estimates of the probability to get it plus-minus the standard deviation times a certain coefficient (anyone has a quick tool to calculate these for binomial distribution? I doubt that the normal-distribution estimate, where `a` should be about 2 for a 95% confidence interval, is suitable here, especially for the smaller sample sizes), which depends on the confidence interval. For each of the five columns respectively, we obtain:

`0.20 ± a*0.07, 0.40 ± a*0.03, 0.55 ± a*0.03, 0.66 ± a*0.04, 0.73 ± a*0.07`

Now, it looks interesting at first, but even within the ‘three sigma’ (a=3) range the probabilities are more or less indistinguishable… And in particle physics, for example, where sample sizes tends to be much greater and the distribution so much closer to normal, they tend to prefer even more, such as 4-5 sigma if memory serves, to claim something definitively… And then again, I’d also appreciate if someone could actually estimate the appropriate confidence intervals for a binomial distribution like here.

‘Who in the Light’s name’, as Cariel would put it, is ‘Rarran’? Anyone can broadcast anything on the platform of their choice nowadays, more or less… I wish people would read books, use their heads, do the research and so on, whatever is appropriate for the situation at hand, rather than give in to their flock instincts and follow The-Light-knows-whom blindly, generally speaking.

I don’t know if the reporter of this data (respect for that, I’m by no means trying to denigrate anyone or their contribution) even knows what a correlation is… But I don’t feel like doing this kind of analysis here, sorry — besides, sorting out that matter of a reasonable value for `a` first would probably be more beneficial from a practical viewpoint.

I would assume that, in order to minimise their workload, the developers would probably adjust the rates (and add another possible events for level-30 Heroic bounties), rather than rewrite the entire algorithm… But then again, it is only a guess.