On Sunday 13 June 2004 21:19, Mark Richards wrote:
Well, assuming you have access to more information than I do, because what has been published is precious flimsy, I'm prepared to take your word for it on JRRT, but what about LT? I see no connection in edits.
In this and future cases, would this formula be acceptable as an evidence?
Wikipedia has n users which have "Troll" in their username, and exists for time t.
Therefore, on average, user with "Troll" in its username is created every t/n.
If an user with "Troll" in it username is created let the T be the time between its creation and banning of the last user with "Troll" in its username.
Then, p=((t/n)/2)/T represents probability that the user is previous user.
If the probability is higher then some value (say 10), ban the user. There's only 10/p chance that you are wrong.
(I am not a statistician but surely someone can check and improve this.)