It's like writing the Anarchist's Cookbook about how to make a bomb, and writing a tweet about how you're going to plant a bomb in Boston in a couple minutes. The latter is a direct threat and requires immediate action. Most of the time, it's too late even for that.
Personally, this dude should've known better and restrain his impulsive internet lulz behavior.
Basic Bayesian approach to probability tells us that this metric of predicting violent events is counter-productive due to a great number of false positives.
Obviously, you've never played any online games.
There's an important point that should be made here.
http://en.wikipedia.org/wiki/Confirmation_bias#Biased_search...
You may have good metrics on the number of incidents that were pre-dated with announcements, but you probably have very bad metrics on the number of announcements that were followed by incidents.
There's no way you can make a statement about how strong the link is without that information.