One-way links? Options

[bwelford]

My K9 spotted the following spam e-mail message this morning.

Hi, 

I am the partnership manager for a group of well known casino websites. I'd like to propose a one-way linking partnership with your website. To explain, I will link 23 of our quality sites to yours if you link to a different 23 sites we own, thus creating a one way link partnership. All our links will be from PR 5/6 sites, and will represent the same quality as we are receiving from you.

If you are interesting in a mutually beneficial relationship, Please submit your site at http: // www .online-casino ...

and you will have 23 links pointing at your site instantly, you don't even have to wait for me to check it. I'll give you 24 hours to link back to us from PR 3+ pages.

I look forward to working with you..

Victoria

I wonder whether in the latest Google frenzy such one-way linking arrangements will be spotted. :?

[rustybrick]

I got the same spam email. lol

Anyway, this has been used for a while now. Ever since before the Florida update, I believe.

The nick names for this one-way linking strategy is called "Pyramid Linking Strategies" or "Triangular Linking Strategies". Now doesn't that make it sound that much better.

How will Google combat this? Are they spotting this? Not sure how they would from a link to link standpoint.

[Black_Knight]

People forget just what a different perspective Google has. I mean, effectively, they have mapped out the entire web, and some of the things people think are way too smart to be spotted actually stand out a mile from that map view.

It usually makes me chuckle.

Let's give this a scenario to help. Imagine two spies who are worried they'll look suspicious if they make an exchange of ducuments in a busy location. So they decide to make a cover of a huge game of pass the parcel - in a busy location. Naturally, this only stands out all the more. The cover may well be more suspicious than a simple exchange would be.

[Paul_H]

Do you think detecting these 3-way links will be built in to the normal ranking algorithm or is it something google only use to manually check for networks etc?

[Black_Knight]

A pocket is a pocket, no matter how many links it takes to go around the circumference of the pocket. I think a certain amount of auto-detection for this has always been in place, in that detecting unnatural linkage relationships is the key to both. I don't think google ever used anything as simple, and flawed, as just checking wither page A links to page B and page B links back. I believe they looked for discrepancies in the natural 'balance' of link structures.

The web is a huge network, and though you can place a group inside that network, and network the sub-group together, it still needs to be connected thoroughly into the big network around it not if it is not to stand out.

[altyfc]

I've got several copy of that same email too (on addresses which take the format webmaster@ or info@<someofourdomains.com> - not addresses we've ever publicised, but we got them because they were on a 'catch all'). Looks like "Victoria" has been very busy... :roll:

Aaron

[rustybrick]

Ammon, good point. But it does make it harder to locate. And its never as simple as site A to site B and Site B to site C and site C to site A. Throw a few more variables into the loop and it makes the map a bit fuzzier.

The question is, how much effort is Google throwing at the detection of link trading? How much are they doing this currently, and what do you think they will be doing in the future?

[Black_Knight]

But it does make it harder to locate

Actually, I'm not sure that it does. It depends on the perspective or tools one has. Most people assume looking at it from the browser perspective. From the 'web mapping' perspective things would look very different.

You don't have to look at whether site A has links to site B which links to site C or so on... that's the browser perspective. Instead you could test a 'six degrees of separation' type thing from a half dozen central points of the web. That's just one simple way of eliminating 'pocket networks' from the equation. No matter how many jumps of pseudo-sites the links went through, they'd never be connected to Kevin Bacon so to speak.

[rustybrick]

Right. I *see* what your talking about.

Normally you should see a web mapping of sites within the same communities, that would be natural.

But if you see a strong linkage between sites in communities outside of each other, that can raise a flag.

Is this what you are talking about?

[Black_Knight]

Ever wondered what applying a bayesian filter protocol to link mapping might do? There'd be 'good' linking structures and 'bad' linking structures, and sites would be scored for both, so not only would you need to be just six degrees separated from the 'good' parts of the net, but you'd also have to ensure that you weren't somehow only six degrees from a bad neighbourhood.

Would your link-relationship map look like the structure for a true point of reference site, or like one that had artificially inflated its linkage with others?

See what I mean about the perspectives making a big difference here? Google have a map that no-one else has. What that map shows them is something the rest of us can only speculate upon, but if anything, most folks assume way too much simplicity.

A spider doesn't walk around counting links. Think more of broad-view linkage snapshots, almost like doing a traceroute with something like SamSpade

[rustybrick]

Yea, that is exactly what I was visualizing when you posted first.

Ammon, nice posts as always!

[Paul_H]

Very interesting posts.

Wonder if and how google use this in their ranking algo. Guess mapping the links is relatively easy but i imagine for an algo to decide what’s unnatural would be harder. Would this be theming?

is there any sign Google currently look for unnatural linking? Would reciprocal linking look more natural then buying links on unrelated sites?

[Black_Knight]

It would be a good move to add theme or topical analysis to the mix, but I'm not sure what the additional resources to do that would be. That it doesn't have to be worked out in real time, and can instead go back to using the same kinds of data used by IBM's CLEVER project could easily allow it to be topic sensitive (equivalent to theme detecting).

However, the root of it is classic old hubs and authorities stuff. Take a good long look at the work by Kleinberg and see how the system identifies hubs and authorities. This is almost certainly the core principle that would be used to analize and 'value' link relationships.

To locate new topical hubs, you start by looking at the known authority sites, perhaps only as many as a dozen or so would be needed for this purpose, and you then look at the backlinks of those known top authorities and find the sites that link to a majority of them. The more of those known authorities are linked to by any single page, the better a hub that page is for the topic.

To find new authorities, you look at the known hub pages and follow their links.

This is why genuine outbound links can be so important, and why the spate of 'directory listing' pages scraped from existing hubs (such as other directories and SERPs) have been so prevalent in the last couple of years.

[clarkhaddock]

What if all of my link partners have my link on their site pointing back to site A( my main site that I want to promote and get popularity to). Then I have their links place on site B (which has no link or connection to my site A), what then? Would the SE's/Google even know what I am doing? After all the whole point of link exchange programs(most peoples) is less about letting people see your link on their someones site and clicking it to come to his but it is more about link popularity with the SE. How many links are out there. I would think this system would work well and Google would not be able to tell what is going on.

One thing that I would not do is have a link from site B to mine so i am still confused on how Google would know what I am doing. In the thread you talked about they has a triagle or sometimes even more points than that which were linked. My idea is to have a triangle but not have it completed/linked all the way around. In theory, the exchange sites are linked to site A and my site B is linked to them but site A and B are not linked. Can google still tell?

[kensplace]

doh I thought this was a doctor who related post

[Black_Knight]

Think patterns, Clark. Especially unusual patterns. Yes, there'd still be an unusual pattern that would stand out like a sore thumb the moment the trouble was taken to look.

[Ruud]

Isn't the "related:" search basically the beginning or foundation of a way to find unnatural links? We're looking there at a filtered result of two sets of pages, resulting in two lists which are diff-ed against eachother to come up with a common structure. That according to the patent apparently behind it they accumulate the differences between the two end lists as a seperate third list could be helpful to them as well.

By looking at pages linked to from the pages pointing to your page, and then doing the same for *those* pages in my idea helps them already "skip" once to get a bird's eye view. It would make sense that a flower shop and a legal site both link to a certain site "A" in the middle -- but what would be the unifying factor in these two sites, each again related to totally different sets, if they in turn were linking to eachother?

Much of this seems pretty abstract. If you're a visual person doing several searches using the Touchgraph Google Browser could be an eye opener.

Thing is ... in the end we know they can spot, oh, say cloaking... but they do not persue it as aggresively as they suggest or some believe. Wouldn't, and doesn't, the same then hold true for any unnatural linking?

[lucadg]

Hello everybody,
I think you made it quite clear but let me be sure of it: a network of natural links is good, right? Not spam, right?

I have about 10 accommodation websites in Eastern Europe, all real stuff with real unique content. I want to create a network of sites in order to cover all the cities in East Europe. By doing this I give my customer a list of selected partners for his needs in the cities where I can't help them. One site, per city, carefully selected based on quality, not PR.
Every site in the network hosts a php Network Table containing links to all the sites in the network.
That means: if you join my network you'll have, say, 20 outbund links and 20 sites linking to you. Same as the scheme quoted at the beginning.
Furthermore: there's one central site where every site in the network can show 4 accommodations.
It's all clean, real stuff. Will search engines consider this spam in any way?