I wasn’t raging at all, but you obviously still don’t get the point. The values you take aren’t any representative, as a single person could easily push the representation of his own spec a lot higher than 10 other guy of another spec could potentially do in that week. That is what my example as pointing at.
No, I just showed you that your assumption is just based on a random number. If I change that random number, I get completely different results. So how representative can your assumption be?
Oh boy, that is my daily life. I am earning my money with it!! You are the one not doing it right! You take games played with a spec but then break it down onto spec count in the ladder. That can’t work as the ladder is player based, but the popularity is based on games played!
If you want to compare both, you would need to have the data of how many people did play the spec last week, not of how many games were played with the spec. You say in the number of games played all that effects would equal out, but that is just another assumption. So your final assumption is based on another assumptions!!
But not enough with that, you don’t do it really spec wise. You said UDK has a representation of 16.53% but that is only based on melee specs. That means, you would need to compare the amount of DKs in the top 1000 of all melee players, not the top 1000 of all players regardless of the spec.
If you just want to take the top 1000, you need to take the overall “popularity” of the spec, compared with all other specs, regardless their role (melee, range, healer)!
And like already mentioned, do I get completely shifting results when I change the observation range. While top 1400 would say that dk is doing a tiny bit too good, the top 200 player suddenly say it’s not doing good enough and need buffs! How could that potentially be, if your statistical survey had any proven relevance? It can’t and that would prove that the significance level is too low to make such a meaningful statement.
You also have not taken into account the changing factor of the ladder and popularity. First of all you took the EU popularity from xunamate, not the NA popularity. If you then compare these with the NA ladder, that would be more than grossly negligent! In NA, UDK had a popularity of 17.33% and DH 20.65% for example, based on melee specs only of course.
And not mentioned so far is that the popularity is based on the last week, but the ladder is up to date. This means that you are comparing historical data with a snapshot in order to draw conclusions.
That’s why I said your time was wasted as your work simply does not stand up to any scrutiny!