[graeme]

This script linked to in the article is very interesting. It shows random and non-random distributions side by side:

http://bl.ocks.org/roryokane/4358325

I used to play Catan a lot with a group of engineers. I noticed that even their mathematical intuitions of randomness were off. For instance, one would try to shuffle the deck so that Knight cards were spaced evenly apart. They did it in the name of randomness, even though their deck manipulation was the opposite of random.

[praptak]

There are some situations where it makes sense to label data as "too random". Cooked books, election fraud and steganography can sometimes be detected by checking for atypically high randomness levels.

In this context it night make sense to make a game less random as it might make it more realistic.

Also: Wheeeee, I'm in the article :-)