Forum Moderators: open
I think they are applying some kind of KWD penalty, my top sites had ODed on the keyword density. The only top ranking I still have is a page with sensible KWD.
Where have your old top ten listings gone to, are they in the 20s now ? I have just had a further look and I now have a new theory that in order to be in the top twenty, the page must have changed since the last round of indexing.
"Never change a well ranking page" might explain why the top twenty have almost all disapeared on masse.
Here is something to chew on.
[alltheweb.com...]
The search was made for the Swedish word "turistbyråer" which means tourist offices. Fast has language recognition and knows what language was used. Observe listing #4. It is certainly relevant to the search, BUT, the page does not contain the keyword. Nor does the rest of the site. It contains a near equivalent in English.
So how did that page get there? Clever cloaking? Perhaps, but somehow I doubt it. Could it have something to do with themes or fishing results out of a directory? Can't think of any other explanations.
And while you are at it, try to figure out how #7 made it into the search reply at all. Totally baffling.
>the page must have changed since the last round of indexing.
Certainly not the case with me Mark.
I'm looking at the domain name and the text in links both internal and external as being more important.
More opinions welcome!
Truth be told, until today I didn't have any top tens on Fast -- I figured I was just a klutz at optimizing for them and turned my energy to the engines I had a handle on.
BTW, I just checked a relatively inactive client. Their index page went from 56 to number 3, and the content hasn't changed in 6 months.
After further research I am now convinced that a change in the page is a major factor in determining its ranking, and the top twenty rankings disapearing seems more than just a coincidence to me.
If you compare your listings in alltheweb and allthesites do the descriptions differ ? I have noticed some top 20 rankings in the new listings where the alltheweb has truncated the description slightly from the old listing. If this qualifies the page as having changed, then my theory holds true for all of the searches that I have tried so far.
Can you give me a search term that contradicts what I have said.
"Nearly all my sites which were sitting in #1 positions have now all moved to #21, the #2s are at #22 etc."
Ditto, exactly the same here.
My pages which were in the top 5 have all moved down 18 - 22 places. This has happened accross the board. Some of the pages have different keyword densities so I dont think that KD is it.
Mark, I like your idea of pages which have not altered being the ones which have dropped. It makes sense and having had a look - it looks like this is definitely the case.
For one of our local engines, changing some of the content regularly results in pages staying up in the rankings and not dropping. I have got around this by adding a content rotator to the pages. Will do this for alltheweb and see where it gets me.
NFFC -
"I'm looking at the domain name and the text in links both internal and external as being more important.
More opinions welcome!"
Always felt that text in internal and external links have always been important in regard to Fast. But, looking at the profile of pages doing well, definitely more now then ever.
By the by - fasts crawlers is in one of the sites which were dropped right now hitting the pages it dropped.
This update has hurt me big time!
Try your search on google and look at the cached entry for towd.com [google.com].
Google says the page has incoming links with turistbyråer in the text.
So as NFFC said earlier link text is important, particularly incoming link text. With this also known to be true in google it is perhaps no surprise that the results returned by both engines are similar.
Number 7 is for a very similar reason, but I will leave the answer to that one hanging for the time being :)
OK. So now we have one element of Fast's new algo. Incoming link text has had its weight increased. And by quite a lot if a page totally void of the keyword can make it to page one. From this I think it is fair to deduce that they are intentionally chasing Google (and its potential licensees).
Well spotted Mark. Thanks.
Thanks for confirming that. You are right KWD is not the major factor. You might not have to have changed your page to be in the top 20 (still waiting to see evidence to the contary), but if you were in the top 20(*see note) and you haven't changed you drop out - guaranteed.
I too have lost heavily on this round, but am pretty confident on getting back up next time. I actually changed my pages well over a month ago, but obviously not in time for this database update.
*note - 20 seems to be the cutoff point, but I have seen 18 used on some less competetive phrases.
To answer renke's question from above see here [ipa-sweden.m.se] - Interesting .
Has anyone else found any other strange looking results that may give us further clues to the algo?
I am still not convinced that the algo has changed. It could just be a new DB and some of the off page criteria having more effect.
Now back to how I was positioned before they dropped me. I was ranked number 2 or 3 for keywords that were only contained in my title and description tags. They were nowhere else on the page.
To Mark:
I think you have a point with the theory on not updating your page within the last index gets your page dropped. The site that was number one before and is still number one now has daily updates to their index page.
Could they be doing something similar to what Excite has done in the past? Where if you do not update your page before each spider visit you get dropped.
I don't know but I'm going to start updating more often.
Ahem... In the search I linked to above the last update for #1, 5 and 9 were as follows:
#1 Nov 24, 1999
#5 Nov 07, 2000
#9 July 26, 2000
I would say, that the result in that search is based mainly on link popularity in combination with link text, with additional account being taken of title and perhaps keyword density. In the example, Fast has managed to present correctly - and in the right order - the two sites most relevant to the search. Both sites have lots of links pointing to them containing the keyword in the text. #1 more so than #2.
I think there has been a change of algo and in fact Mikkel Svendsen - Fast's biggest licensee after Lycos - hinted at that a month ago in another thread here at WebmasterWorld.
>To answer renke's question from above see here [ipa-sweden.m.se] - Interesting .
Mark: ????? I have a very small brain and don't understand what you mean. What I see is a poor links page in Swedish. And.....
Renke from the search for turistbyråer I see:
Old New Page Changed
#1 #2 Yes
#2 #1 Yes
#3 #8 No
#4 #9 No
#5 #10 No
#6 #11 No
.. .. ..
#25 #30 No
So on this search, if you haven't changed you drop to position 8 or lower. This fits in with the note I added above (about 20 not being the absolute cut off), this search returns far fewer results so the drop is less.
The swedish links page contains a link to unitedstates.com as well as the word turistbyråer. The word is not in the link text, but close by. The next question to answer is does the word need to be "near" the link or just on the page ?
Fast licencees Allthesites.com, Evreka.com and Lycos Sympatico.com show exactly the same result for my test for "turistbyråer". Instead of Meta description they show the first text from the 1st paragraph of text. According to startup, this would be the old db and/or algo.
Alltheweb.com shows a different order where the oldest page is #1 (last updated 1 year ago). That page resides on a site with 1700 external links pointing to it, many of which have the keyword in the text. It beats all the other search replies hands down in that respect. If startup is right, this would be based on the new db and/or algo.
So: if new vs old page matters, it can be easily outweighed by linkage and link text.
[Added] Mark I get your point re link text vs. proximity. Now that is food for some very deep thought. Do you think they might have gotten hold of Brett's Web Page Reputation Calculator?