| 5:30 am on Apr 18, 2007 (gmt 0)|
|I was also wondering if the 'penalty' is applied more often to sites with more than one page on the topic? |
EG wikipedia has 4, one of which ranks *very* poorly considering the site it is on.
So... all the sites with news about widgets are doomed?
If this was the case the 900 range should be full of blogs.
| 5:42 am on Apr 18, 2007 (gmt 0)|
What happens when the 900 range is full? Will they expand it to the 800s or just start dumping sites/pages completley?
| 10:47 am on Apr 18, 2007 (gmt 0)|
You must search for only non-competitive things.
To overgeneralize, if any particular site seems to be targeted for the 950 area it is all those hacked .edu sites... which fits in with the most annoying aspect of the problem described here: normally high scoring authority domains being put in the 950 box.
The 950 penalty hits edu's way more disproprotionately than other tld's, but that is just because basically by definition there aren't a lot of spam pages on highly respected domain... and there are now thousands/millions of such pages on edu domains.
So, in the warped Googleworld, penalizing a middling amount of pages on respected .com sites "logically" follows. They are LOOKING for the hacked edu pages. They are trying to penalize those, but not the entire edu domain.
If you think they are carrying this over to the rest of the web, the phenomenon makes a bit more sense.
| 10:52 am on Apr 18, 2007 (gmt 0)|
"Will they expand it to the 800s or just start dumping sites/pages completley?"
Both. The "950 area" can be fairly big, but aside from that, tens of thousands of -950 penalized pages are discarded. Obviously if there are ten hacked edu pages around 950 for some search, there are likely many thousands of others that are likewise penalized but not showing.
I suppose the highest scoring ones are shown, which makes sense again in why we see there solid-PR authority sites that normally rank well.
| 12:04 pm on Apr 18, 2007 (gmt 0)|
They might not dump them, they might just turn them into supplemental pages.
Or they could dump them, who knows
| 12:17 pm on Apr 18, 2007 (gmt 0)|
Here is a thought, the algo thinks a page is spam, turns the page into -950 so that human reviewers can look at them easily?
One thing about this thread, if it indeed was a google error we would have seen matt or adam post to this thread saying ok, we see we have an issue and we will work on it.
Since they have been tight lipped on this and have not commented, you know what that means. The algo is not broke, these results are borderline spam.
| 12:35 pm on Apr 18, 2007 (gmt 0)|
|One thing about this thread, if it indeed was a google error we would have seen matt or adam post to this thread saying ok, we see we have an issue and we will work on it. |
What drives you to that conclusion? Has there ever been any time in history that a Googler posted saying "We have f*cked up?" Not in over 7 years that I can recall, but if you know of such a post, a link to it would be much appreciated.
|Since they have been tight lipped on this and have not commented, you know what that means. The algo is not broke, these results are borderline spam. |
Again - what drives you to that conclusion? And how do you know what that means - if anything at all?
| 12:44 pm on Apr 18, 2007 (gmt 0)|
Remember Big Daddy when sites were going into the supplemental index?
Googleguy recognized the issue and looked into the error and fixed things.
When the -30 penalty came up, tedster asked about it and recieved a big no comment from google durning pubcon I believe. When penalties come around, google is very quiet about them for the most part. If there is an error, typically we see a post from adam or googleguy saying they will look into it. They read these threads on webmaster world.
| 1:11 pm on Apr 18, 2007 (gmt 0)|
this -950 is no error, its a filter thats 100%, the trouble just still what activate that filter, I also thin there can be many aspects, not only links, keywords....
| 1:37 pm on Apr 18, 2007 (gmt 0)|
Yep, it is a filter. Easier to demote you to #30 than giving you a white toolbar. Also you cannot whine that you have been booted from the index.
Clever eh? To be honest we overdid our links to our penalised site but who doesn't these days? All the other sites in travel have mega linking. Only way you can get to #1 for city hotels these days is have more backlinks than the next guy.
| 2:16 pm on Apr 18, 2007 (gmt 0)|
The more posts I read, the more I do research, the more I'm convinced.
TrustRank "became thematic" in its effect.
It's tied to the main themes of relevance and exact phrase matches.
Add the co-occurrance filter ( too many themes colliding ), and you can mix all kind of "penalties" and name them at will.
Links pages with lots of text and money phrases that the page is not trusted for - are busted.
Innermost pages of big-big names without proper relevance from the navigation ( or no relevance from "home" to pass with the navigation ) busted. Sometimes for using too broad anchor text. Sometimes for using too many popular, only marginally relevant phrases. ( Too many co-occurrencies of themes that the page is not directly relevant for, are within the same sector, but are seen as their own category based on search popularity and advertiser competition. )
Collateral damage is... any otherwise legit website with a navigation that doesn't have a proper theme pyramid supported by its anchor text / inbounds. Reaching EVERY page on-topic. AND/OR pages that mass up too many themes.
That "computer science" search...
Chose something obvious, the umass.edu group.
It's about computer science allright, has it in the title and all... is linked to from a lot of other universities and its own subdomain root as well.
I'm lazy so I only checked the first three dozen anchors.
There wasn't a single inbound with "computer science".
Their internal navigation uses "CS" for short too.
And Google doesn't know that CS is short for Computer Science.
That travel site too... ( my-city hotels ) with the domain name and all incoming anchor text using the exact query. If you take away these words, and their combinations, this is what the internal navigation looks like:
|Find a Rental Car |
Other Theme Parks
Travel Destination Sites
New York JFK
...of which neither the page, nor the entire subdomain has inbounds for. Their entire network was based on the idea to pass trust and pagerank from each and every subdomain to the other, crosslinking pages that were only relevant to a single city name and service.
My-City Hotels was voted from pages relevant for Other-City Hotels. My-City Car Rental from My-City Hotels.
Most are highly competitive, monitored closely.
They're their own category and can't be "used in parts".
What this means is:
My-City ( generic search ) is one theme. Informational.
My-City HOTELS is ANOTHER. Travel sector.
If you're top 600 for My-City, you could narrow it down ( or expand? ) to different sectors. But that's a lot of trust. A lot more than for a two word query.
If you're relevant for My-City Hotels you can't add an anchor using My-City Car Rental, News, Attractions... etc, unless you have support for these themes.
If you had a travel destination site sitting at the top of the relevance chain ( which this my-cityhotels.whatever.tld does not have ) supporting each subsection with a theme pyramid, you could do this:
Main Page ( My-City, generic )
Subsections ( Travel sector, Informational, Retail, Whatever )
Innermost pages linked from ONLY the appropiate subsection
( travel --> Hotels, resorts, airports )
( news --> weather, local news )
If you don't crosslink different themes excessively, you're safe.
But in general, with thematic TrustRank, this is essentially spam. We all know the links to other KIND of services/information and other cities could have been added as dropdown menus, for example. The reason why they are lined up by the side, with the exact search queries as the anchor text, on ALL pages is...
Not so long ago TrustRank was a generic score that you either had or did not have. You decided what you'd use it for.
But with these additions, not anymore.
Also to add, hotel sites ( and other such too competitive themes ) spam trustrank hubs almost habitually but they rarely/NEVER can find a source that'd be relevant to them. This now includes that, even if on-page factors of the source are a match, if it doesn't have any relevance passed from the hub root to match the theme, then in the end it's NOT relevant ( see example at the top ). I've seen so many instances of trusted, but completely off-topic pages linking out to hotel chains ( and flight companies ) in an obviously inappropiate way, that them losing all kinds of deemed-irrelevant inbounds during the past few weeks wouldn't surprise me at all.
Ever wondered why only high trustrank sites sit at -950?
This is exaclty for what we think it is for.
Trusted but "irrelevant" pages are -950. How come? Because they passed the threshold, however, the on the fly reranking puts them behind everyone else for various problems. The rest is not -950, but OUT. This filter may apply to all sites, but it's definately TrustRank that's KEEPING some of them in the primary index at all. Trust is not lost.
Also, I think this applies to all pages and links, not only trusted sites. It's basically a way to battle "off topic" links and pages, from links bought only for their parameters ( PageRank ) to link exchanges of the same nature.
I'm only using this abstract to address the problem so that former "authorities" who have high trust can relate to the issue. This is a filter that not even high TrustRank makes a site invulnerble, that's all.
I know this is is not scientifically correct.
It's over simplified.
Because I'm interested in ( predicting ) the behavioral pattern of the algo, and not mapping its anatomy.
I don't have time for that.
This is taking long enough as it is already.
[edited by: Miamacs at 2:28 pm (utc) on April 18, 2007]
| 3:33 pm on Apr 18, 2007 (gmt 0)|
Its also a very stealthy and slow filter as well. If you do not pay attention to your pages, you will probally see them slowly drifting into the supplemental index as some members commented earlier on this thread.
I imagine that trust rank goes up for a site over time depending on their linking out/in. I am more than sure google monitors links over time.
So its not impossible to come out of it, you just have to really be careful with links out to sites, recip links and paid links because it can get you into trouble. I would not worry about the scrapers trying to link to your sites. Many sites in the top 10 have a ton of scrapers linking to them and they do not show issues.
| 3:44 pm on Apr 18, 2007 (gmt 0)|
The site I mentioned that went -950 on the 10th came blazing back last night with higher rankings than before.
I went through it with the proverbial fine tooth comb, stripped down each page to the bare minimum of keywords and because of the possibility of a phrase based indexing component in the algo also removed most of the related phrases.
So now the keyword phrase is just in the title, meta description and a few times naturally on the page. I also removed the H1 tags just to really strip it down to the bare bones.
I also had an inconsistent navigation, on 20 percent of the pages it was different. So i standardised that across the site.
So it came back big time, but perhaps if I'd have left it alone the same would have happened? Plus it could very well happen again in days, weeks, months. But I am hearing a lot of folks talking about deoptimising pages having a positive affect.
The site is not involved in any link exchanges, it's an authority site linked to by many others in the field and a few .edus, newspapers etc.
If you think your site may be borderline over-optimised try out the motoricerca Search engine SPAM detector and see if it can pick up anything dodgy on your pages. Also go through your site with xenu, make sure there are no broken links or orphaned pages that google thinks are doorways.
Or you could just leave it alone and hope for the best. Sometimes that works just as well :)
| 3:51 pm on Apr 18, 2007 (gmt 0)|
As mentioned before, mine disappeared on the 10th - I've not touched it yet and it's still down there. I've not experienced any bounce in and outs as discussed either. I plan on deoptimisation and link strengthening but if that's to increace trust from authority domains I may just be compounding the problem as laid out by miamacs.
[edited by: LineOfSight at 3:54 pm (utc) on April 18, 2007]
| 4:10 pm on Apr 18, 2007 (gmt 0)|
|I also removed the H1 tags |
Were you using any other heading tags or just the H1
| 4:44 pm on Apr 18, 2007 (gmt 0)|
Just the H1. I think once a site builds up a certain level of trust you can rank on just on in bounds, title and subtle on page keywords in content....Anything else could push you into the filter because you are already themed and maybe Google thinks that borderline spam.
If you stay shackled at -950 for more than a month I would say go for deoptimisation and see if you can break free.
But consider if you already have high msn/yahoo rankings messing with your site could knock those down.
But if google is your main referrer and you are currently getting hardly any traffic from them what have you got to lose by deoptimising? You can slowly reintroduce later.
| 4:46 pm on Apr 18, 2007 (gmt 0)|
Great post Miamacs
| 9:16 pm on Apr 18, 2007 (gmt 0)|
jk3210, I certainly would not call that a competitive search.
Instead its a good example of high volume, low competitiveness. Searches like that attract mega-volume of similar template-type pages that are just thrown up for zillions of localities. Google's mission on sorting through that is different than in more competitive areas.
If the edu hackers aren't targeting something, almost by definition it isn't all that profitable.
Suppose Google has decided 2% of all edu pages are hacked, so they identify the spammiest appearing 2% of the pages on any domain that has an authority score above "X". (Spammiest based on too much or too little anchor variation, too many synonyms, too sticky, too high a clcik rate compared to surrounding results... whatever, many possibilities.)
Suppose Google then decised to penalize all those 2% pages, because since they have such a lame search engine, if they did not the alternative would happen to be that they would have the top 100 be 90% results from hacked edu's. So they choose the lesser evil.
Except for the fact that this doesn't effect every site, that seems to be what is going on in most cases... with the added aspect that Google's judgement is quite bad regarding what are the 2% most spammiest pages on any authority domain.
| 5:16 am on Apr 19, 2007 (gmt 0)|
|what I usually find. Lots of Yahoo, TA, and hotel chains |
| 1:32 pm on Apr 19, 2007 (gmt 0)|
I found some of my pages in a cluster in the 950 region when searching a term to see if another page is in the index yet. I immediately searched for each of the 3 pages and found they rank well for their natural search terms. I sighed a great sigh of relief but now I'm wondering why they are down there. Could it be they are a clue to the 950 cause?
| 4:17 pm on Apr 19, 2007 (gmt 0)|
Searching A+B+C you're on top.
Searching C+D+E the same page goes to the 900 range.
It's like Google thinks that the same page is spam for certain keywords. Very easy to fix for them, very hard for us.
| 5:04 pm on Apr 19, 2007 (gmt 0)|
An error is not the same thing as an "intended consequence" (quote Eric Schmidt), but the results are so widespread it might just a well be an error.
I sure like all the fancy theorizing, but I just wish that we'd admit that SEOs don't know what Google is up to anymore.
[edited by: Martin40 at 5:37 pm (utc) on April 19, 2007]
| 11:55 am on Apr 20, 2007 (gmt 0)|
|you just have to really be careful with links out to sites, recip links and paid links because it can get you into trouble. |
Google's take on paid links has been made crystal clear, but as far as reciprocal links are concerned, you say they are bad and Adam Lasnik says they're fine done in mderation.
So please now tell us - who should we believe and listen to, you or Adam?
| 12:30 pm on Apr 20, 2007 (gmt 0)|
I understand paid links, but tell me the difference between link exchanges and recip links? I will give you a link, you give me a link to gain some PR! Now a lot of these people way over due it! One link, might be ok, but a lot of sites do it in the 50's, 100's and 1000's which is pure abuse and being involved in linking schemes.
One member here who was caught in this -950 penalty had over 2000 "recip" links. Come on, thats PR manipulation.
| 12:46 pm on Apr 20, 2007 (gmt 0)|
Also, like I said earlier do some research on number one results. I would say 99% of those site do NOT have this type of page:
mysite.com/whatever with these words or something to this nature wrote on them:
If you would like to trade/swap/link to us please use this code "blah, blah, blah" and paste it on your webpage. We will then verify it and place a link to your site on our page as well.....
Come on, I am sure google can pick up these kinds of phrases and flag the page as link spam. Especially if the page is useless to searchers and has hundreds of recip links.
| 12:58 pm on Apr 20, 2007 (gmt 0)|
Trinorthlighting, I appreciate your enthusiastic evangelism, but this is not normally the form of "reciprocal linking" which is used by high ranking sites, even if you could define "reciprocal linking" in the first place.
People tend to like and cling to black-and-white answers and if I have learned only one thing during my time here it is that the world that we move in has all kinds of shades of grey...
| 1:07 pm on Apr 20, 2007 (gmt 0)|
I think BillyS said it best:
For what it's worth, here is my policy:
- If you've got good content and I'm referring to that content in my content, you get a link.
- If I'm writing an article and I consider your website an "authority" on the topic then you get a link.
- If you want to advertise on my website then you get a no-follow tag.
- Send me an email about exchanging links and you go right in the trash.
If you follow that advice, you will never find your site being flagged for linkspam.
| 9:32 pm on Apr 20, 2007 (gmt 0)|
That is of course utter nonsense, and completely off topic.
| 9:49 pm on Apr 20, 2007 (gmt 0)|
Problem is it's not that long ago that reciprocal/exchanged links were being lumped together with bought links by both Google and MC.
IMO there's a difference between reciprocal and exchanged links, the former can be accidental, the latter not.
Would Adam endorse exchanged links?
| 11:44 pm on Apr 20, 2007 (gmt 0)|
I have a site that was hit with this late Jan, I got my site out and am still out after about a month, before I post what I did I would like to look a few affected sites and try the theory on those, I am looking for sites that have been hit site wide not individual pages, anybody interested PM me, I cannot promise anything and will only give advice the work will have to be done yourself.
| 12:00 am on Apr 21, 2007 (gmt 0)|
as many has said here, we see sites, with H1, heavy link trades, lots of keywords and much more at the top and this is what we have talked about could be wrong with those sites in 950 filter, but how about this:
You are a adsense member, you have adsense on all your sites, you have more sites be cause you need a safty net, if google plays around so at least one or 2 sites still rank and you can pay your bills, but Google uses there adsense to see which sites you own, if you own more in the same category, then only 1 of those sites rank with all pages and it dossent matter if the one is about all widgets and the other is specialy about blue widgets.
| This 194 message thread spans 7 pages: < < 194 ( 1 2 3 4  6 7 ) > > |