View Single Post
Old Oct 1, 16, 1:36 pm
A FlyerTalk Posting Legend
Join Date: Jul 2001
Location: Watchlisted by the prejudiced, en route to purgatory
Programs: Just Say No to Fleecing and Blacklisting
Posts: 91,902
Originally Posted by pa3lsvt View Post
Binomial Distribution

This distribution applies when the probability of success ("being selected") is the same for each independent trial. Like I said earlier, when the probability changes from event to event, the math is more arduous but can be calculated.

A common fallacy among non-statistician types is that the presence of unlikely events is evidence of a non-random selection process - yet Powerball has winners when the odds of winning are greater than 180,000,000:1. Except for the few examples where the selection process was manipulated by cheaters, lottery ball systems are designed to be truly random.

There is also the fallacy of dependent trials. Just because you were selected last time, and the time before that, and the time before that, and the time before that (however unlikely) does nothing to make selection on the next trip any more or less likely.

And to be really accurate, the selection algorithm is not truly random at all but is likely based on a seeded pseudorandom structure of some sort. Hopefully the WTMD algorithm is statistically random, but I'm sure that info is SSI.
The presence of unlikely events may be evidence of a non-random selection process. And it's the presence of unlikely events that helps some to determine if the selection process is random or not truly so.
GUWonder is offline