18 replies to this topic

[Black_Knight]

Before anything else, let me make perfectly clear that what follows in my post is a hypothesis, not even proven enough to be a proper theory as yet. Indeed, my main purpose in posting this is to see if anyone can offer hard evidence of any kind to test the validity of this hypothesis.

The Google penalty (PR0) and automated penalty hypothesis

In this hypothesis I believe that there are at least two separate forms of penalty behind the infamous PR0 penalties.

One is hand-applied, and is the type applied to the PRadnetwork and also to SearchKing from what Google have admitted. These kinds of penalties are never coincidental and are the classic 'bad neighbourhoods' that you can be penalized for linking to. Sites hand-selected by Google are, by definition, the very core of bad neighbourhoods.

The second form of penalty is automated, it is not so much a penalty as a part of the algorithms, and is based upon linking to a bad neighbourhood. This is the type that hit many of SearchKing's portal partners. This is what makes a bad neighbourhood so apt a term, because there is guilt by association. If you link to a site with the hand-penalty, some of that penalty may transfer to your own site. If that happens, anyone linking to your page may also inherit a small portion of the penalty.

Now normally that automated penalty has all kinds of limits. It won't trigger on just one bad link, and it tapers off quite sharply in transferrence. However, by the very nature of dense cross-linking (as seen in link exchanging networks), when one site has some automated penalty, it goes around the other sites and creates a virtual feedback loop. Each site just keeps making that small penalty cycle around (through the reiterative PR calculations) until it drags them all down.

Remember, of course, that the penalty comes in from outbound links, not inbound ones. This is to ensure that no webmaster could create a domain that deliberately earnt a penalty and then deliberately spread that penalty to others by linking to them. The penalty can only come by linking to sites that are penalized, not from them linking to you.

This is how I am so certain whenever I see a penalty that outbound links will be at the heart of it.

Want to see how it spreads? Look no further than the Zeus collective. At least half of the sites in the collective have managed to infect each other simply by linking around. Its the feedback effect that does it. One bad link will hardly affect you, but if you link to two other sites that also link to that one bad one, or just heavily to each other, then the loop has begun.

Now Google don't really consider the automated penalty as a penalty because it isn't a black-list. In fact, they may have no way of telling who has been hit by it. It would be considered (as it is) an integral part of the ranking process.

Page A links to Page B
PageRank goes from A-B
PenaltyRank goes from B-A

That, in nutshell, is my hypothesis.

I've had this hypothesis in my mind for some time, and you may find that an older discussion at WebMasterWorld has some relevance for further consideration.
WMW: Who you link to and possible penalties

[Grumpus]

This is a good theory. I like the "feedback" concept as I believe that some penalties are really just a little bit of a penalty (say, enough to bring you from PR5 to 4 - or even a High5 to a Low5). But, when it gets in that loop (I call them "dead-ends" or "pockets" where sites link only amongst themselves but have no way back OUT to the rest of the web) that penalty keeps applying to itself over and over again until it hits 0.

The more sites you interlink with, the more times that feedback can come back at you. That's one of the reasons I've always been fond of just linking to good sites on a whim rather than expecting an exchange - it's a good outlet to get the bots out to a good area in the event that some inadvertant bad link does exist and you haven't caught it, yet.

I also think (although this likely isn't true for a site/page that's been PR0'd by hand) that a link out to an automatically penalized site is not nearly as bad as a reciprocal link between you and them. If your site links to them, but theirs doesn't link back, the feedback loop can't get started. (That's a hypothesis, though, not even a real theory).

Good thoughts, BK. I agree with you almost completely.

G.

[peter_d]

Its the feedback effect that does it. One bad link will hardly affect you, but if you link to two other sites that also link to that one bad one, or just heavily to each other, then the loop has begun

Nice. It certainly sounds plausable, and mirrors the way PR works.

Did you read Googles patent application? If my understanding of that is correct, Google looks to build a dataset of related hubs, and if a link to your site features in enough of them, you get a boost.

This could also work in reverse. Google looks to a dataset of penalized sites (negative hubs), and if a pattern emerges whereby you appear to share a lot of links with them, you're out. So it's not just PR0 feeback, it's also hub/authority related.

Pure speculation

[Shurlee]

Sounds close to the most uniformed hypothesis I've ever heard but if you're right, then the whole PR, google algorithm thing is in real trouble. So are all of we.

http://www.alltheweb...hking.com&c=web

If you're even close to right, that is one big bad neighberhood! We all better go through all those 45,000 + links and see who we might be linking to or who might be linking to us.

What are you suggesting we watch out for when trying to link the "right" , (google), way. Should we stay away from PR 4's like the searchking site as well as
http://www.aim-pro.com/contacts.html or should we just be careful we don't ever link to any site that has the words search or king in it?

What do I do about about linking to the AIM's site today to find out, (hopefully), later that they were stupid enough to p***o** google and NOW I better not be linking to them.

Personally, I think google is smarter than that and there is damn little that happens that they weren't expecting to happen. I also think your post does little to help anyone figure out penalties and how to effectively avoid them.

It does have a certain, "google is just trying to help us all out and keep the internet clean", ring to it though.

[aspwiz]

Maybe you can run sites that you link to through a search on google....
enter the line below exactly as is, replacing the domains for the right ones

site:www.yourdomain.com "+www.siteyoulinkto.com"

If you get results then take your links down.

[Shurlee]

Slow day today, (boss is not trying to look up my skirt so I thought I would do a little research on the hypothesis.

This page
http://www.alltheweb...uk &_sb_lang=en
show the #1 link as a PR0
#2 a PR 3
#3. a link to a google SERP, (where do ooutgoing g links come in?)
#4 PR 3
#5 PR 4 (YIKES!)
#6 PR 0
#7 PR 5 (happens to be this forum)
#8 PR 0 (my hypothesis -- LINK FARM)
#9 PR 4 (Yikes !)
#10 ANOTHER PR 4

I wanted to go on but it's not like it's my day off. So, according to your hypothesis, why aren't you a PR 0

I'm pretty sure you'll now explain how I got it all wrong. Just curious though.

(As a side note, just in case you're right, I pulled my links to some of the SK lawsuit stuff down.)

[aspwiz]

I should have pointed out that it only points out results with the TEXT of the domain.....

The text, may in fact just be text and not a link.

THIS ->>> searchking.com is just text, and not a link and certainly unpunishable.

[peter_d]

I also think your post does little to help anyone figure out penalties and how to effectively avoid them. 

Speculation can be as much a basis for discussion as anything else.

Google has said they dislike link farms, so it is only logical they look to penalise them, by algo and by manual editing. If by algo, then how is it done?

[DaveChild]

I wanted to go on but it's not like it's my day off. So, according to your hypothesis, why aren't you a PR 0

Because he's not linking to them. PR can't be raised or lowered by someone linking to you, you have to link to them. What Ammon is saying is that linking to a site with a penalty will hurt your ranking, but when that becomes especially effective is when there's a group of sites all linking pretty much exclusively to each other - then the penalty just keeps going round and round the group until everyone hits rock bottom. (I think ... not sure I've interpreted what Ammon said exactly though).

[Black_Knight]

Remember, of course, that the penalty comes in from outbound links, not inbound ones. This is to ensure that no webmaster could create a domain that deliberately earnt a penalty and then deliberately spread that penalty to others by linking to them. The penalty can only come by linking to sites that are penalized, not from them linking to you.

As you can see, ILJD is right in his answer to Shurlee's question.

Mel, I think the above answers your thoughts too, but if not, please help me to see just what it is you're getting at, and I can then tell if this strengthens or weakens the hypothesis.

Shurlee, I can't control who links to me, so won't be punished for that. It is the links I control, the outbound links, that could get me into trouble. This part is well known of course. What I'm trying to get to is how come a good page can link to a core 'bad neighbourhood' and get away with it, while other sites may have just a few links to sites that link to bad neighbourhoods and get PR0.

My hypothesis is the feedback part. That the reciprocation is the point, in that all pages in the circle could get a tiny part of a 'PenaltyRank' which would do nothing on its own, but as the links cycle, a feedback loop is developped so that all pages cross-linked get infected with a small share of this 'PenaltyRank', and just keep on infecting and reinfecting each other, a cumulative poisoning with each iteration of the PageRanking process.

[Black_Knight]

Mel, thanks for the clarification.

The only way I can see that happening with normal PageRank calcs is if the bad links were set to some minus value, because no matter how many times you add zero to a link that is giving you a PR4 rank the sum is still 4.

That's exactly what I'm getting at Mel. However, it can't be in the normal PageRank calculations because they work exclusively on inbound links, and the penalty works exclusively on outbound links.

I'm thinking it has to be something along the lines of a damping factor, but one that builds cumulatively rather than just removes a proportion of PageRank. It would (IMO) be an anti-PR, a negative value that could completely counter other positives.

We're on the same wavelength here, just there isn't yet a set of terms for exactly what we're getting at.

[Grumpus]

I have nothing to back this up, but it is possible that every page has two natural factors (leaving content and all of the other stuff alone) - PR and PF (PageRank and Penalty Factor). A page will normally have a PF of 0, but if it links out to a bad page then that PF may up to 0.00212. Then, any page that links to that page gets a factor of that PF passed backwards to it. Get caught in a loop due to crosslinking and/or other factors and that 0.00212 could become 2.12 before long. If your page is a "natural" PR2 due to inbound links, the "penalty factor" would make it a PR 0.12 or 0.

If this is the case, then I'd suspect that it works exactly the same as PR. If the only link on the page is to a bad neighborhood, you're going to get a massive percentage of that penalty back. If you have 50 links on your page and 2 or 3 of them go to penalized pages, but the rest of the pages are good, then it's split up a bit and you don't even know you've got a penalty because it only brought you down one fraction of a PR point.

<shrug> most of that was train of thought speaking, so it may not be worded very well. I think it's an interesting extension of Ammon's theory to explore, though. </shrug>

G.

[DaveChild]

...PR can't be raised or lowered by someone linking to you, you have to link to them

Ammon I was referring to this statement, which simply is not correct, PR comes from sites linking to you, not from sites you link to, with the exception of earning a PR0, of course.

Sorry about that - I do know PR comes from sites linking to you, but must have been having a dozy morning. I meant "PR can't be lowered by someone linking to you..." etc.

[aspwiz]

Surely if a *bad neighbour* were to link to you, it would have *some* effect??

[Grumpus]

Not on PR. The only effect that a link to you from a bad site has is no (or reduced) benefit from the link. It has no negative effects whatsoever. If it did, then there would be all sorts of sites that became bad neighborhoods on purpose just to stomp out competitors.

G.