JoeSinkwitz - 4:18 am on May 11, 2007 (gmt 0)

I agree that this is probably part of it, but like you say...not likely all of it, or at the very least, not in its full incarnation (otherwise we probably wouldn't be complaining).

According to the patent the normal non-spam doc has something like 8-20 related phrases, with spam docs being above 100, but in examination of EOS sites the 8-20 phrase rule is far closer to what is going on, so that figure is being thrown out of whack.

A funny way to think about things is that if Google messes up on the first time through the ranking, re-ranking will only make it worse, since relevant sites will be thrown out. A ton of super-authority non-relevant sites (like newspapers and edus) can drastically alter the expected phrase scores.

Thanks for posting the patent again though...it was helpful to re-read. I'm going to drastically de-optimize a few sites, since it'll probably bring them back, and then start readding the text users want to see piece by piece [assuming some of the longer lasting EOS issues ever leave the spam table...some I've seen do back and forth, some just once a month for a few days, and some having come out since Jan].

Cygnus