« Blog Clog Continued... | Main | Massive Bandwidth Savings proven! (Empirical Evidence of RFC3229+feed savings...) »

September 30, 2004



Some people have far too much time...


That only counts links in actual posts, right? So it would miss sidebar or blogroll links. Some of Zeldman's most influential work exists not on his own domain, but on Digital Web or A List Apart.

Bob Wyman

Matt, Linkranks are, as you suggest, based *only* on links from actual posts (items in RSS or entries in Atom). We do not consider links from blogrolls or other "HTML" links that aren't in posts. If you are interested in numbers that do consider such non-post links, I think the numbers that Technorati generates would include them.

bob wyman

Greg Gershman

A link rank would be relative thing, right? It's not an objective gauge of your popularity, just a comparison of your popularity versus others'. As the blogosphere grows and evolves, it is likely that technology blogs will lose some of their former prominence, as more non-techies start blogging. It's possible Zeldman has the same average number of links, or possibly more, than he did in the past.

Response on 3-October-2004:

Greg: LinkRanks are, in fact, relative. Thus, they are measures of popularity -- not absolute numbers. That's why we call them "ranks" not "scores."

We compute a "score" internally and then rank order sites according to score but don't publish the scores themselves. (Although we're considering it.) Because of the ranking function, some information is lost for the sake of making inter-site comparisions easier. For instance, the difference in score between sites that are only a few ranks apart can be very large for those with low LinkRanks but very small for those with high LinkRanks. You can get a hint of this by noticing that the variability in LinkRank is much lower for those sites with low LinkRanks then it is for those with higher LinkRank. For instance, Amazon has always had LinkRank #1 except for on one day when it traded places with nytimes.com which normally has LinkRank #2. On the other hand, if you look at sites with higher LinkRanks you'll see that they move up and down in LinkRank much more often. (Compare the "bottom" of the top 100 to the "top". You'll see many more green and red numbers, indicating movement, among the numbers at the bottom of the list.)

As for Zeldman's numbers, what you're basically suggesting is that while Zeldman may have weak relative strength compared to the entire universe of sites, he might still have good relative strength compared to other sites in his industry or topical segment. Yes, that is possible. However, it is not the case at this time. The drop in LinkRank for zeldman.com, in this specific case, is too great to be explained in this way. This would be easier to show if we were computing "sector-based" LinkRanks -- which we hope to do in the future. However, we've wanted to get a bit more experience with the LinkRanks themselves before making the system more complex. If you have any ideas how we might be able to improve the system, please let me know.

bob wyman

The comments to this entry are closed.