Recently, the good folks over at NovelRank released the results of an interesting study they undertook in order to determine how various sales events effect book rankings on Amazon. Given the recent hullabaloo concerning the "fake your rankings" ebook scandal (if you can call it that), this comes at an appropriate moment, and gives us some real and useful information rather than the "scam your way to being famous" variety we've been inundated with of late.
The criteria of the study essentially set out to disprove (or support) several common "myths" concerning how Amazon's ranking system works. Amazon won't divulge the information themselves, but the practical application of a few controlled experimental purchases provided some useful facts nonetheless. Among the most interesting discoveries was that returns do not negatively impact sales rank (i.e. the book is returned, but the sales rank remains), and that sales of a book on multiple orders increases the ranking more than multiple sales of that title on the same order (so if you're buying your own book, only buy one at a time!).
Visit the NovelRank blog post to read the details and learn more. Also, if you're an author I encourage you to employ their definitive ranking and analysis tools to your benefit. They're very cool, they're impressively fast and thorough, and best of all, they're free!
One "myth" not tackled in the study, however, was that "if you buy your way to the top, someone else will buy your book too."