Whoa - 7:59 pm on May 16, 2011 (gmt 0)

It's a great list, Brett. Nonetheless, I tend to agree with some of the folks who have said that 1) this type of list is going to be used by Google more actively in the future, but that 2) Panda is not using this type of input extensively now.

If these factors were being used by Panda, we wouldn't see such bad results in Google now, as somebody mentioned before in this thread. There are lots of sites that intuitively warrant high bounce rates for their bad content that have benefited from Panda. So, it's hard to believe that engagement metrics are at the heart of the new algo.

To me, more than anything else, it seems like they just dialed up their reliance on PageRank, relying on it more so than they had in the past. To develop Panda, they got their sample sets of good and bad sites; then they ran their indices through a wide array of new algo factor weightings (tweaking weightings on old factors and trying some new ones as well); and, after all of this testing and simulation, for whichever weighting set gave the best, tightest correlation on their good versus bad sample set, they annointed that set of weightings to be king.

They were probably very happy to see that a greater emphasis on PageRank was in the winning set. Unfortunately, that change has messed up many of us. A PageRank 5 site with a million pages could do very well when on-page and other non-PageRank factors were weighted heavily; once Google upped its emphasis on PageRank, those sites suddenly didn't have enough PageRank fuel in the tank to continue to rank well. In the new era, the relevance of your page doesn't matter nearly as much. This theory would also account for Google's comments that one bad section of your site can bring down your rankings for pages anywhere on your site. If you have many weak pages without any links, your strongest pages are going to perform poorly regardless of how many links they get. Anyway, that's my theory du jour. Thanks again for the great post.