Having tired of listening to arguments about what goes on at AFD but having little data, I decided to engage in a little experiment to analyze voting patterns at AFD. I wrote a computer script to parse over 100 days of AFD logs and compile a variety of statistics by looking at the bold faced terms (eg. keep, delete, merge, etc).
Obviously this is imperfect since AFD is suppose to be about discussion and consensus building (somehow despite somehow having 100 noms and 875 votes cast a day), but it is good enough to reveal some interesting patterns. For example, I would not have guessed that on a forum with 7200 participants over 100 days that the top 5 would account for 10% of all votes. Or that for all articles with 2/3 of the votes cast favoring deletion, only 6 would also have 10 or more keep votes.
However, I am not all that surprised that regulars cast delete votes 76% of the time whereas occasional participants (the vast majority of all voters) say delete only 56% of the time.
The full listing of statistics compiled so far can be seen at:
http://en.wikipedia.org/wiki/Wikipedia:AFD_100_days
Things are still somewhat preliminary as I may yet try to further improve the parser and the last three days of my AFD sample have not entirely closed, but I doubt those things with substantially change the patterns.
Anyway, I hope this provides a better grounding for some of the endless discussions about what to do with AFD (and whether or not it has serious problems).
-Dragons flight
PS. I need to get back to real life now, so if there are questions/requests/complaints, it will probably be several hours or a day before I get back to respond to them.