Forum Moderators: open
It doesn't appear that there will ever again be a Google "update" in the traditional sense of the word, or even by recent criteria for calling something an update. I personally don't believe there's any such thing that'll ever happen again, not even when the backlinks and TPR "appear" to be updated. IMHO that's just showing something that's already happened, with no telling when it happened.
I also don't completely believe that what's being called "sandboxing" is necessarily a probationary period for new sites or links, in spite of the apparent delay that's being consistently experienced. I think it's got more to do with some kind of off-line processing so that the processing real-time can be more efficient than it otherwise would be.
Where in the world would I get such a hair-brained, half-baked theory? From reading Kleinberg - SETS for starters. There are a lot of later papers related to links-based algorithms, but this very early one I came across makes a few points that are interesting for today
Automatic resource compilation by analyzing hyperlink structure and associated text [decweb.ethz.ch]
Very interesting indeed.
What stands out to me,is that via the combination of authorities and the 50 byte text proximity thing, is that this really could be a way of programmitically diminishing the value of links from pages with multiple off topic themes often distinguished as 'sponsored links' or 'adverts'.
Who wants to go to a 'hub' page? If I wanted to do that I'd use the DMOZ not a search engine like google.
And who apart from webmasters builds a web page with links in the text which describe authority pages? Almost every website ever just has a links page or at least a menu with no descriptive text at all.
Nice idea, now try the real world.
[news.cornell.edu...]
That's the very early work, way back when Infoseek was still around, and Yahoo rather than ODP mentioned as a resource. So many that have come after place stress on topical relevance of pages linking - and that's aside from PR, though it's been expected for a long time that we'll see some flavor of topic sensitive Page Rank sooner or later.
[edited by: Marcia at 2:54 pm (utc) on July 15, 2004]
Thanks for sharing your research with us, Marcia
Re-examination of algorithmic implications is always a good idea.
There are three kinds of updates from Google, algorithmic, SERPs, and PR. The Florida Update was algorithmic, and widely believed to be the result of Google's inclusion of Hilltop, wherein the anchor text of links was given much greater relevance to a page's score as related to the search term used, whereas pre-Florida SERPs we largely scored by link-pop and on-page presense of search terms. I think we will continue to see algorithmic updates, but probably only once a year or so, simply because significant innovation in algorithms is so limited.
SERPs updates are brought about by re-crunching the web itself. As google has announced, these are on a rolling basis and now seem to be fairly undetectable in that there's no major change in the landscape such as can be brought about by algorithmic change. PR updates seem to be the last bastion of reliable excitement from google, inconsequential as they seem to be, they still provide an indication that the wheels of change are spinning at google.
My personal hope is that what's been unfortunately come to be known as the "sandboxing effect" is the result of offline crunching and not knob tweaking from the mysterious Oz-like google wizard behind the curtain. If it is the result of knob-tweaking, it's likely a time filter intended to offset google-bombing. If it is algorithmic, it is likely the result of offline crunching of a search term list against the whole of the google indexed web to establish hubs and authorities.
I have noticed that google has taken a tip from Microsoft, which often labels bugs as new features. This is apparent in that stemming is actually a side-effect of LSI, and we all remember how google promoted stemming as a new feature, never telling us it was actually the result of implementing LSI.
I do think we are in for if not due for another algorithmic update, perhaps on the scale of Florida, and that this will be the expansion of the current semantic indexing to include more than just the anchor text on inbound links, such things as the Title and/or h1 components, and maybe even semantic scoring from the site as a whole. From the way people already clamour about getting backlinks only from other on-topic related sites, you would think this had already been implemented. Perhaps the clamour is merely the reflection of herd expectations and wishful thinking that will never come to pass, or perhaps it could happen next week. If such data were to become part of link scoring, it could be implemented in Hilltop, LSI, or other, any combination, or all, and if done all at once could cause a fundemental restructering of the google index even more dramatic than Florida.
