Print-on-demand web site, Lulu.com has a service that helps project how well your book will do based solely on its title. The algorithm was developed exclusively for Lulu by statisticians who studied 50 years' worth of bestseller data encompassing over 700 titles from 1954 to 2004. They compared the attributes of these titles against a control group of less successful titles by the same authors which gave them a model to project the odds of any title becoming a New York Times bestseller.
For example, here are a few trials I ran…
Catcher in the Rye – 63.7% chance of being a NYT bestseller
Twilight – 63.7%
The Da Vinci Code – 35.9%
Pride and Prejudice and Zombies – 22.9%
Outliers: The Story of Success – 10.2% (I think the algorithm only looked at Fiction bestsellers but there's no reason you can't try it with non-fiction as well – though obviously, this result indicates that there may be some flaws!)
(I heard about this site in a book I just finished called “Super Crunchers” whose author actually used a Google AdWords campaign to help choose between two alternate titles for his book!)