Google: November Algo Update-dance, Discuss all current ranking issues here Options

[Alan Perkins]

Also, on-page factors don't mean much in terms of ranking. You simply can't tweak a page to the nth degree and rank anywhere for, say, i am a spammer. On-page has only ever worked where there is little or no competition.

As there are many different definitions of SEO generally, there are also many different definitions of on-page SEO. In my definition, for example, it includes copywriting. So, using my definition, on-page SEO is still important.

So it shouldn't matter if the site is filled with hidden text and other nonsense, so long as it passes the authority test then it's place is deserved. The existence of text, hidden or otherwise, is a non-issue - the on-page factors aren't the reason for the ranking.

I disagree.

I think we agree that copywriting matters. (IMG:http://www.highrankings.com/forum/style_emoticons/default/smile.gif) Well, IMO hidden text is hidden copywriting - so it matters, too.

Also, a page builds authority from other pages. Off-the-page often means on-another-page (i.e. elsewhere-on-the-Web), particularly in the case of Google. So Google will have to continue to watch for this nonsense - in particular, hidden links - not only to see if you're trying to deceptively improve the ranking of your own page, but also to see if you're trying to deceptively improve the ranking/authority of another page.

[Peter]

I wonder if the average searcher notices anything different at all? My point- I don't think searchers are going to abandon Google in droves.

That is very true Scotty,.. I have the same experience and when I ask anybody that isn't into rankings, they tell me every time (no exceptions) they haven't noticed anything. I have to explain really well why I asked.

Peter

[dannysullivan]

The idea there is that they pull a preliminary set of search results, and then calculate this local rank score from the interlinking within that group of pages. The problem with implementing it in "real time" is that this involves a lot of additional computation.

Yes, but this is exactly how Teoma operates. It calculates its own form of LocalRank at query time. Teoma's remarked to me in the past on how much computation this involves and why they've felt they've had an advantage over others who might want to do this system. It's hard, and others couldn't do it.

Well, I'm sure Google could do LocalRank if they want to. Moreover, I think they are doing this. I think that when you run any standard query, you're getting back LocalRank results. However, when you run a difficult query -- such as one that involves excluding a word -- then Google falls back to its old system.

If I'm correct, that explains why you see old site coming back when you exclude a word. In fact, it doesn't have to be a made up word. Go do an advanced search, then run your query and exclude XLS files. You'll see the filter "override" in action.

Running a query for ALL terms -- what Google does by default -- is less hard then running a query for SOME terms and NOT others. For those type of queries, doing LocalRank and exclusion may just be too hard for the moment.

No doubt there are other things going on in the changes that people are speculating on. However, I (and others) see changes happening even for non-commercial terms, and there are plenty of optimized, commercial sites that are still ranking well with Google. It could be they have a predefined list of term and/or sites they are aggressively filtering in this release. But it could also be that they're simply using two radically different ranking systems now.

[ibizwiz]

scottiecl realistically points out that most searchers are still (probably) going to find what they want in Google, so

> I don't think searchers are going to abandon Google in droves.

I don't either, Scottie. But all evidence at this point is premature, since no one to my knowledge (except, maybe, Google or MSN) is studying the behavior of searchers in a rigorous way.

Think about it - for TV and the net itself we have organizations like Netratings carefully measuring what folks are doing. Here is online search, perhaps the fastest growing industry in history, and we have nothing yet except a few weak studies with woefully small sample sizes.

For our team, understanding how searchers interact with their search tools is our single highest priority, but I doubt we can say much publicly or definitively for at least another year or two. <sigh>

Since all we have is personal and anecdotal evidence, let me add that I spent yesterday destroying two large turkeys with a group of people who included four active search users - none of whom are "in the business". Two *had* noticed a sharp fall-off in quality in Google, at least in the topics they were interested in. In both cases the primary complaint was commercial results (in the SERs, not the AdWords column) that simply dumped them in big ecommerc sites, where they would then need to try and search for their query targets. Indeed, one (71 years old) had noticed the site reported in the NY Times that is using multiple doorway pages in violation of Google practices. He was very annoyed.

So, while I agree that Google's user base will not erode quickly, I conclude that it is NOT because folks will not pretty quickly see or sense the escalating commercialism of Google, but rather because they simply do not see an alternative right now.

Given the huge potential revenues at stake, we can be sure that, as alternatives emerge, their developers will use SEM and other highly targeted, web-based or opt-in email-based methods to attract disgruntled Google users.

As many others have pointed out, search users are famously fickle. If they have even a single "search experience" that they perceive as particularly satisfying, quickly productive, or, my guess, simply more fun that the last few searches they performed using Google, the seed of doubt will be planted, even if an outright switch does not occur then and there.

[OldWelshGuy]

The filter (or filters) appear to be 'selectively applied' , but i can't help wondering, as Danny has what exactly is driving the filter. It is after all an algo and as such will have a numeric cut off point at some stage, balanced with processing time/load maybe? or maybe just by straight volume of returned pages.

maybe it would work like this:-

search for xxxxx check database>yes this is a value added search term as decided by adwords toolbar history etc. goto search> results less than xxx's do not apply filter, results more than xxx's apply filter, search term containing any of the known localisms, ie UK Wales England Scotland Ireland or any other dbsed term then apply local rank filter and as the results would be low and the search results be so specific there would be no need to attack them with the 'highly targeted' filter.

Maybe I am off target here I have just read it back twice and I am not sure if i can understand me, so lord knows what chance other have (IMG:http://www.highrankings.com/forum/style_emoticons/default/wink.gif) but i am tired, and it has been making me think too much, plus I am old Bald & stupid lol

[Peter]

However, when you run a difficult query -- such as one that involves excluding a word -- then Google falls back to its old system.

These types of searches are in no way more difficult than a normal search. All these type of operators influence the selection process, not the ranking process. I think practically every type of command, like allinurl or allinanchor, etc. influence the selection process only. The ranking process is the same for all of them.

The selection process is quick, the ranking process is what takes time and computational power.

The local ranking algorithm, if in fact implemented, wouldn't have such a huge influence. Local Ranking influences the top 100 results or so. It re-ranks those pages based on the links between those 100 sites. So just the order would change, it wouldn't introduce other pages into the top 100.

The logic behind local rank is that the top 100 (or 50, what ever they chose) is very likely to be a community, and there for has many links between them. Within a community there is probably a number of them that are considered the leaders and this shows through the links between the pages.

Computing local rank between just 100 sites wouldn't take that much time and there for could very well be implemented already.

Regards,

Peter

[Peter]

The filter (or filters) appear to be 'selectively applied'

It is my impression that this "selectively applied" filtering is pretty random. (Not all commercial sites are effected and not all very popular or competitive phrases are effected, and also not in the same level)

Would you consider the fact that perhaps a different way of ranking with different factors given different importance, could look like "selectively applied" filtering?

Regards,

Peter

[OldWelshGuy]

QUOTE
perhaps a different way of ranking with different factors given different importance, could look like "selectively applied" filtering?

Yes it is very possible all will be revealed over the next few months lol

[samt]

So it shouldn't matter if the site is filled with hidden text and other nonsense, so long as it passes the authority test then it's place is deserved. The existence of text, hidden or otherwise, is a non-issue - the on-page factors aren't the reason for the ranking.

Oh I have to thoroughly disagree with that. I have a certain competitor who I recently discovered was using hidden text, stuffing keywords and other people's company names all over his pages, in the same color as the background. This competitor and I, we are linked in the same directories, from the same trade associations, etc. It was always kind of a head-scratcher to me, how the guy ranked well, because his content (that I could see), was exactly a copy of the manufacturers' sales literature, making his pages appear to have identical text to a whole bunch of other pages out there. Obviously, he wasn't tagged as duplicate content, becuase of all of the hidden text. I'm not worried about the guy, but obviously, when competition is close and off-page factors are pretty equal, this hidden text does affect things.

[Peter]

Yes it is very possible all will be revealed over the next few months lol

Well everybody just calls it a filter all the time,.. I just don't see why,.. there is no proof of that.

ah well,.. doesn't really matter.

(IMG:http://www.highrankings.com/forum/style_emoticons/default/cheers.gif)

Peter

[Denyse]

I found this report Bayesian Spam Filtering

Has anyone seen this - and what do you all think of THAT theory - sounds pretty convincing to me.

[powerofeyes]

Well everybody just calls it a filter all the time,.. I just don't see why,.. there is no proof of that.

If you want to see whether it is a filter or not, Just add the -junkword to the search and compare the search to the normal search. The sites which are ranking now were once in 40's and above. Since most of the above sites are knocked off those sites are holding the top place now. That is why many call this a filter, If you want just have a check,
VIJAY.

[Think Web]

I found this report Bayesian Spam Filtering

Has anyone seen this - and what do you all think of THAT theory - sounds pretty convincing to me.

Interesting...

Well, we've all been asking the Google godz to interceed on our behalf to cleans their SERP's of spam.

Have they accomplished their goals? We'll have to wait and see.

[jbelle]

I found this report Bayesian Spam Filtering

Has anyone seen this - and what do you all think of THAT theory - sounds pretty convincing to me.

I was very curious about this theory after finding it via the Slashdot discussion on Google's recent changes ("Google Blocks 'Optimized' Pages") but I'm confused as to how Google intends to make Bayesian spam filtering work when--from my limited understanding of it--it really works best at a user level. I guess this is why Seth Finkelstein sees the results of Google's filtering as having "unintended consequences"? Additionally, if one filter is being applied accross the board (like a server-level filter on email, as opposed to user-level), spammers need only crack that one filter to make it through.

Hmmm... Would Google (or perhaps the next big engine) begin customizing for each user by constantly requesting and processing data from individuals regarding how well each search went, how accurate their results were, etc. thus creating an individualized search engine with personalized results?

[marcuse]

Well everybody just calls it a filter all the time,.. I just don't see why,.. there is no proof of that.

ah well,.. doesn't really matter.

Oh, it matters...


We are going to do a test on two words to see whether there is a filter in
place.


The first word is "lingerie".

Current Google results:


www.trashy.com/ - 2k - Cached - Similar pages
www.cameolingerie.com/ - 13k - 27 Nov 2003 - Cached - Similar pages
www.2cutelingerie.com/ - 12k - 27 Nov 2003 - Cached - Similar pages
www.victoriassecret.com/ - 14k - 27 Nov 2003 - Cached - Similar pages
www.threewisheslingerie.com/ - 9k - 27 Nov 2003 - Cached - Similar pages
www.allhot.com/ - 14k - 27 Nov 2003 - Cached - Similar pages
www.lingerie.com/ - 10k - Cached - Similar pages
www.angelbodywear.com/ - 31k - 27 Nov 2003 - Cached - Similar pages
lingerieatlarge.com/ - 16k - 27 Nov 2003 - Cached - Similar pages
www.adultlingerieuk.com/ - 8k - Cached - Similar pages

Lingerie -googlegoo results:

www.lovefifi.com/ - 57k - Cached - Similar pages
greentithelingerie.com/ - 26k - 27 Nov 2003 - Cached - Similar pages
www.victoriassecret.com/ - 14k - 27 Nov 2003 - Cached - Similar pages
www.lingerie-of-hollywood.com/ - 16k - 27 Nov 2003 - Cached - Similar pages
www.2cutelingerie.com/ - 12k - 27 Nov 2003 - Cached - Similar pages
www.trashy.com/ - 2k - Cached - Similar pages
www.cameolingerie.com/ - 13k - 27 Nov 2003 - Cached - Similar pages
lingerieatlarge.com/ - 16k - 27 Nov 2003 - Cached - Similar pages
www.feelingsexylingerie.com/ - 8k - Cached - Similar pages
www.angelbodywear.com/ - 31k - 27 Nov 2003 - Cached - Similar pages



Now do you see the difference between the two searches, normal
and -googlegoo?




Now let's do the same for unfiltered "jacket":

Normal google results:

jacketmagazine.com/ - 26k - 27 Nov 2003 - Cached - Similar pages
www.mymorningjacket.com/ - 3k - 27 Nov 2003 - Cached - Similar pages
www.bluejacketdrama.com/ - 8k - 27 Nov 2003 - Cached - Similar pages
www.no-contact.com/ - 12k - Cached - Similar pages
www.wired.com/news/technology/0,1282,58914,00.html - 25k - Cached - Similar
www.jacket.org/
www.amazon.com/exec/obidos/tg/ detail/-/B00005ATQF?v=glance - 74k - Cached -
www.imdb.com/title/tt0093058/ - 48k - Cached - Similar pages
www.krug.org/scripts/fullmetal.html - 101k - Cached - Similar pages
www.islandrecords.com/ - 39k - Cached - Similar pages

Now jacket -googlegoo results:

jacketmagazine.com/ - 26k - 27 Nov 2003 - Cached - Similar pages
www.mymorningjacket.com/ - 3k - 27 Nov 2003 - Cached - Similar pages
www.bluejacketdrama.com/ - 8k - 27 Nov 2003 - Cached - Similar pages
www.no-contact.com/ - 12k - Cached - Similar pages
www.wired.com/news/technology/0,1282,58914,00.html - 25k - Cached - Similar
www.jacket.org/
www.amazon.com/exec/obidos/tg/ detail/-/B00005ATQF?v=glance - 74k - Cached -

www.imdb.com/title/tt0093058/ - 48k - Cached - Similar pages
www.krug.org/scripts/fullmetal.html - 101k - Cached - Similar pages
www.islandrecords.com/ - 39k - Cached - Similar pages


NOW THE CRITICAL POINT: THE SEARCH RESULTS FOR JACKET WITH NORMAL SEARCH
AND -GOOGLEGO ARE *IDENTICAL* YET WITH LINGERIE THEY ARE COMPLETELY DIFFERENT.
WHY? BECAUSE THERE IS A FILTER IN PLACE FOR LINGERIE AND NONE FOR JACKET!
Googlegoo has nothing to do with last month's results but has to do with
demonstrating the existence of a filter for CERTAIN COMMERCIAL KEYWORDS....

Now the challenge to all debunkers of this theory:


Demonstrate non-competitive or non-popular searches which are SUBJECT to the
filter (i.e., differences between -googlegoo and normal google searches).

This is what is called PROOF as opposed to all the speculation around here...