Page is a not externally linkable
Scarecrow - 8:54 pm on Jul 20, 2004 (gmt 0)
My theory is that Google has overhauled their backlink/anchor text index in an effort to purge spam. I don't know if it is a real-time purge, or if the entire backlink index was rearranged in "precomputed" fashion, like they used to calculate PageRank, or if what we're seeing is a combination of the two. The fact that the link command changed so drastically suggests that a good portion of it was precomputed. Here's what it looks like to me: 1) Most people do not put three words into the search box. If you organize a purge of backlink anchor-text "juice" for links that contain at least three words, then most people will not notice any significant change in the results for their search terms. 2) Do a word frequency count for all anchor text in all backlinks. 3) Look at all anchor text that has three or more words, and multiply the frequencies for each word. Establish a threshold for further investigation. 4) Examine the the link targets for anchors that exceed the threshold. Determine how focused the targets are. In other words, how many total pages or sites do they link to? If the focus is narrow, then assume that this is a case of anchor-text over-optimization. 5) Zap these links. Don't take the pages out of the index, but deny those links on those pages the capacity to add ranking juice to their targets. I don't have any proof, but in the thread mentioned above, the link command for Google's target page, using the old and new link command, shows how my Google bomb may have been destroyed. The old link command in Google showed 9 links from my sites with the anchor text "out-of-touch executives." The new link command shows 2 of my links. The reason I believe that three or more words are needed before this new anchor-text algo kicks in, is because my bomb behaves normally if I substitute a different word for the word "executives." The juice is still there, but not for my particular combination of keywords in my anchor text. The whole phenomenon might even be a lot simpler than what I describe above. How about if you downgrade anchor-text linking juice for all links with more than two words of anchor text? Would this explain what we are seeing with the new link command results?
I have a preliminary theory about the new backlink phenomenon. To understand the genisis of this theory, you have to read this other thread [webmasterworld.com].