Category : SEO Bee Tree
At the end of 2003, the Google search engine started using a new algorithm for ranking its search results - the Hilltop Algorithm. Opinions are varied regarding the exact time when this new algorithm was put to use (whether during the Florida Google Dance or before), but there is a general agreement that it is already in use
The Meaning of the Hilltop Algorithm
The Hilltop algorithm was created by Krishna Baharat from California between the years 1999-2000. On January 2001, he patented a certain algorithm in partnership with Google. The patent is different from the one presented in the original article, and it is not clear which of the two options is the one that is actually used. Today, Krishna works for Google. Although Google never officially announced that they are using one of these algorithms, it seems that they couldn't overlook their advantages.
If in the past Google was interested in the number and the quality of incoming links to a certain page, and not in the question where these links came from, today the game has changed significantly.
There is a new webpage status called Expert Documents. In case an expert document links to a webpage, this serves as a "vote of confidence" for the target page, and as a result this page's rank increases in the search results. There will be different expert documents for different search words, when the major difference between Krishna's article and the registered patent is in the way expert documents are selected.
A page that didn't receive a link (a vote of confidence) from at least two expert documents, will not receive a score from the algorithm at all! Although it is possible that it will still appear in search results because of other factors (e.g.: PageRank and factors of the page itself), but the webpage's location will be clearly damaged.
The key difference between the original article and the registered patent is in the way an expert document is selected.
Pre-defined expert document list
If you read the article written by Krishna Baharat (you can read the original Hilltop article here), it looks like the search that is based on the new algorithm consists of three steps:
Locating expert documents: Constructing a pre-defined expert documents list from available webpages. This list is general and relatively permanent. Next, linked sites are removed from the list (see below)
Correlating expert documents with the search query: When a search is executed, it goes through the expert document list, thus creating a sub-list from the extensive expert documents list for the same topic.
Assigning a LocalScore to the search: A LocalScore is given to any page that comes up in the search according to the anchor text of incoming links from pages that are included in the expert document list that was created in the last step. If the page has less than two links from expert documents there, the page would receive no score whatsoever from the algorithm.
The definition of expert documents is pages on a certain topic that were especially created in order to direct visitors to information regarding the chosen topic, meaning that they contain links to other sites that are not affiliated with them. Examples of this kind of pages are: index pages, institutional websites (.org, .gov) and university websites (.edu). The meaning of this fact is that it's not enough to receive links from commercial sites as before. You should receive links from authoritative pages, which are naturally harder to come by.
The importance of correct registration in indexes and other expert websites has increased greatly, while the importance of links from "simple" sites and pages decreased. An additional result is a greater importance of the anchor text of the links that appear on expert documents that link to your site.
Search and ReRank
If you look at the patent that Krishna Baharat registered in cooperation with Google (that can be called the Hilltop Patent), you can see that the search consists of three steps according to the new algorithm:
Initial search: Performing an initial search on all search words (similar to the original pre-algorithm search).
Filtering affiliated pages: Removing affiliated pages (see below).
Assigning a LocalScore: Assigning a LocalScore to pages on the list according to incoming links from pages that are also found on the list. The basic assumption is that the listed pages are the most relevant to this search, therefore only links from these pages will be counted.
A clean up of the expert pages includes removing pages that share the same domain (domain.com, domain.co.il, two.domain.com, www.domain.com) or pages that are located in the same IP group (the first three numbers of the IP number are identical: e.g.: - 212.125.23.XXX).
What about marginal search terms?
Depends on which Hilltop version you believe in. If you refer to the second version, it looks like all it takes is the creation of an initial list that is big enough (after page filtering). If there are not enough websites, the algorithm will not be activated.
If you believe in the first version, then if the search does not appear in the popular search terms for which expert documents were prepared (or found in real time), Google goes back to using the old search method (meaning, without Hilltop).
How do you deal with Hilltop?
First, keep doing what you were doing so far - optimizing your site and building links for PR improvement. The significance of these factors may have decreased, but they are still very significant.
The key to dealing with Hilltop is recognizing the websites and webpages that are defined as experts in your website's field. How do you find them? Depends on which version you believe in. If we are dealing with a pre-defined list, then DMOZ will be one of the experts, and so will Yahoo!. Other indexes may appear as experts, but are not necessarily so.
According to the second version, you need to perform a search with certain keywords, and try to get links from the webpages that appear in the search results. In addition, you should check the incoming links to pages that appear at the top of the search results, and try to get the same links.