It all really depends on how Google values back links, and if back link values can affect each other. Here is some wild speculation on how they might value them and what the effect would be. Back link value = PR*relevance ratio Relevance ratio would be a value between 0 and 1 were 1 equals perfect relevance between the 2 sites Example 1: Valuing back links and totaling them Link1 PR5 * Rev ratio 0.4 = 2.0 Link2 PR2 * Rev Ratio 0.8 = 1.6 Total link value = 2.0 + 1.6 = 3.6 Under this example all back links are good back links, some just give more points then others, but none can reduce your score. No question that this type of calculation is taking place some where in their formula. Example 2: Valuing back links and giving a weighted avg. Link1 PR5 * Rev ratio 0.4 = 2.0 Link2 PR2 * Rev Ratio 0.8 = 1.6 Link Value Ratio = (2.0 + 1.6)/2 = 1.8 Under this example links with low PR or low relevance could actually push your avg rating down. Example 3: Combination of the two methods mentioned above and weighting them against each other. This type of double weighting mechanism is fairly common. Example 1 Link Value * .8 + Example 2 link value ratio * 0.2 2.0 * 0.8 + 1.6 * 0.2 = 1.92 To me this is the most realistic example. By giving a small weight to a weighted ratio versus a totaled value it has a keep em honest factor. Only sites with a really high value would probably be hurt by low PR or irrelvent links.
Libertines, thanks Ccarlow, did you come up with this yourself? Actually, the real formula for PR (just PR, not all the other factors which influence SERP's) can be read in the original paper on the google project. Especially this part: I've written an article which explains this better, see my sig for that.
Yes I came up with it myself, like I said wild speculation. I was not calculating PR though, was calculating how PR and relevance might work together to give backlink values for SERP purposes. I did find the original papers you referenced on the Google project very interesting. Although I'm fairly certain your interpretation of this formula on your site is off. I'm not trying to be rude or argumentative, but I will try and highlight the main problem I see with your interpretation. The PR of a page is not determined by links on the page, but by links to the page and the number of links those linking pages have on them. IE linking to other sites will not lower your PR but will only dilute the PR you pass. Now obviously you put some work into this analysis you are showing on your site, but please point out to me in the formula where it mentions the links on Page A. PR(A) = (1-d) + d (PR(T1)/C(T1) + … + PR(Tn)/C(Tn)) C(A) is not mentioned in the formula, just mentioned in the way of explanation above the formula which I think has probably mislead you and others. In summation the outgoing links on your site will not lower the PR of your site just the amount of PR it pushes to other sites. Edit: While adding more links from your homepage will never lower the rank of the homepage it could dilute the rank of internal pages.
Exactly - That was the one turn off here in the beginning for me, too many people assuming that everyone should already know what they know. Didn't take me long though to say the hell with it and start posting anyway. I just had to learn do jump out of the way when stones and red rep got thrown at me. As or this thread, it was informative for me. I read posts here that say "PR doesn't matter", then I read posts here that say "it does matter". Information like this helps me make my own decision.
CCarlow, I believe there is a misunderstanding, as C(T1), C(T2), ..., C(Tn) is the number of links on the page. I see it's confusing with the C(A) thing in the quote, sorry for that. However, I do agree with it that more links on your page do not lower that pages PR.
One thing that I found interesting that I was not sure if, is that if there is an importance of linking to sites that matches yours. I guess it has some importance from what I read from this thread.
Well I'm not confused but would like to point out that example 5 on your Google PageRank Explained site it clearly describes how linking to external sites can lower your PR. Which apparently we are both saying is not the case. Also every calculation made in the all the examples is off base, made on misinterpretations of the formula.
So this is a nice explanation and discussion of PR in general. However the discussion around "relevance" of inbound links needs to center around, "What does relevant mean?" Sure, as humans it is easy to talk about, but it is Googlebot that gets to decide. So how does she determine relevance? Tman's Quote is interesting. Since I am sure that Google still works on keywords (and/or maybe DMOZ categories for those listed?). Is finding the same keyword(s) on the linkers page and the linkie's page enough to trigger "relevance"? If Google really does more value "relevant" links, then what we really need to do is figure out what makes a linking site "relevant".
If I can go back to TMan's opening post, I'm inclined to interpret the "neighbouring" idea as some form of check in Google's algo to see if (shall we call it?) mysite.com is relevant to the others that show up for the search phrase in question. To use an extreme example, if I do a search for "jokes and puzzles" (no quotes), I'd expect to find a series of sites that focused on just this. At the same time, I would be astonished if the SERPs included something that was dealing with "climbing Mount Everest" (no quotes again). If it did appear, it would have to be humorous story rather than a serious account from one of the climbing team members. And I suspect that this is what Google means. Somehow, they incorporate a check that mysite.com has the same sort of content as the rest of the results. In other words, the "neighbouring" idea probably has nothing at all to do with whether the links from or to mysite.com are relevant or not. Rather, the question is how and whether the collection of sites that turn up for a given search phrase are sufficiently similar in nature to warrant the inclusion of each one of them. Yes? No? Duncan