User blog comment:Flonnaru/Results of Non-Drum Expeditions sparkle effect/@comment-24430017-20150609005754/@comment-24430017-20150609163303

How exactly are you justifying eliminating data as outliers? Are you claiming that some people have different odds of returning GS given the simple condition of 4 sparkled ships?

Not sure what you mean by RNG. Are we talking about true random vs psuedo random? I don't think we will get anywhere useful on that issue. If we are talking about RNG in terms of probability, then yes we do want to measure  "RNG".

If people aren't doing it wrong, I can't really understand the justification for eliminating data. After all all the probabilities should average out. If you do this in math class, you could have all members of class flip a "perfect" coin 10 times and obviously some data is not going to look the same, but you don't check for outliers because there is nothing wrong with the way they did it.

I googled and found this article: http://pareonline.net/getvn.asp?v=9&n=6

It suggests eliminating outliers with a z score of 3.

Sure you can get data that looks "better" by removing outliers. Are we trying to get data that looks "better" or are we trying to figure out the rates involved?