Page rank must die

Surely page rank, by its nature, must die. Especially given that a large amount of page rank is based on links and those sites that have been around longer will simply increase the number of links to a point that new web sites cannot compete. The only way to seriously get any reasonable amount of links is to have teams of people working on it.

Will we not inevitably get to a stage where 10,000 incoming links will be a minimum required for a PR3. Following this logic the whole reason behind page rank is undermined. Instead of people providing links to sites because they thought they were of benefit to thier visitors, it is done to get reciprocals. There can be no logic then to Google's "A link form site a to site b is considered a vote from site a to site b"

Obviously as a new site developer, I will continue to search for links, however I would not be surprised to find reciprocal linking die as a method of increasing page rank and more and more emphasis put on "on site" stuff.

I wonder how many terabytes (or maybe petabytes) are filled by worthless link pages. Or am I just annnoyed because it's now getting to month 3 and still no rank from this very annoying company.

I agree with you. But not so drastically of eliminating PR overnight.

Google became the most used SE because with the internet boom and the big increase of websites, all other SE begun to give less reliable SERP, and the new algorithms proposed by GG were the obvious response to the need of that moment.

Now, as you have said, links between pages is starting to not have the meaning "a vote from page a to page b", but still is a relevant source of information about the quality of destination links. May be the point is, as many other have stated, Content Targeting PR, or maybe GG should restrict the value passed between links, or maybe the answer is in some other place, but the fact is that if SE doesn't improve heavily his indexing algorighms as Internet grown, they are leaving the door opened to some new and smarter SE for appearing and repeat the history once again.

My thoughts go in the sense that one year ago, I can find near everything what I needed at the first page of GG results but now, every new day those results are going less and less relevant for end users searchs.

As the general once said, "Nuts!"

Page Rank was a rather clever but astoundingly naive idea when first developed. Google managed to shoot up like a rocket because PR was, at the moment it was first implemented, a substantially superior method for constructing SERPs.

Because Google's rise was so rapid, it was a while before the naive part became clearer: once you give out the combination to the safe, your valuables are gone. Once a reasonably large community became aware of just how Google was developing its SERPs, "links as votes" became just so much folklore. Excepting perhaps--and even then only perhaps--really small, thoroughly amateur noncommercial sites, links are not "votes"--they are a commodity, bought and sold, often literally for cash.

Google keeps trying to roll out algorithms that will somehow discount this circumstance, which is rather like trying to hide your valuables from thieves by moving the safe from room to room in the house, while leaving its combination unchanged. It does not, it cannot be made to work with any real overall effect.

I don't know who Google thinks it is fooling by dumping some sites here and some sites there off the wagon--mostly themselves, I suppose. Meanwhile a hundred times as many remain on the wagon for the free ride. Most big-time operators now regard domains as disposable: put up the site, run it down and dirty and rake in the PR and the SERPs and the cash till the fat puffing sheriff belatedly catches up with it, then calmly jettison it and start anew with another domain. They do it by the dozens.

It is nowadays, with the advances in computing science (and computing hardware firepower) easier--vastly easier, I suggest--to establish relevance solely by examination of on-page content. Sure, the spammer will try to game that, too, but one page is one page, and can be run through several distinct filters for improper content arrangement. A hundred or a thousand or ten thousand links are, well, a hundred or a thousand or ten thousand problems: there is no straight-forward way, or even any decent clever way, to ascertain the "relevance" of a link, so as to determine if it is an 'honest vote" (as if there were any left)--any bright child can construct numerous scenarios in which apparently "irrelevant" links are perfectly sound. There is no good way to separate the baby from the bathwater, so you throw out both or you keep both. Google keeps the dirty bathwater, because otherwise they'd have scarcely any sites left to list for anything.

We live in an era where, owing to technological progress, size does matter, very much so. It is probably not too hard to make a better automobile than GM does, or print a better newspaper than the one you read--but to enter the market at all takes mind-bogglingly stupendous amounts of capital. Let us say that I am firmly convinced that I could, given the capital, put together a SE company that could easily out-perform Google for relevance, to the extent that the general public could readily tell the difference. Swell.

This is not an enterprise you can start up in your garage and "grow". It is not even one based on some one clever pony trick that you could perhaps do a demo on. It requires a staff of computer search experts developing filtering algos to ascertain relevance from page content, and--perhaps even more important--block spamming of content. It is not something that would be hard for such a staff, but neither is it something that a couple of bright college kids are going to work out on their own.

Nor, considering the low bar that Google sets, are Yahoo or M$ going to bother doing anything terribly new--they'll just try to do a little bit better than Google and rely on marketing to make the difference.

But the day will come . . . .

--
Cordially,
Eric Walker

I believe by the point (or anywhere near) that 10,000 links is required for a PR 3, the internet will be a far different place. Searching will be different most of all.

The growth of the net and online marketing is not moving that fast. IMHO.

Mike

I think PR should become a much smaller part of the google algo.

It is killing new sites and generating a load of crap on the internet.
I cant wait for the day a new engine comes along and takes over the game.

Excellent post!

No real need to quote it verbatim though...


A large amount? Its a link based formular, that is actually called PageRank, no space, capitilasation.

PageRank is great, but it is not Google's algorithm. It is a part of an algorithm in the same way dough is not a pizza, and neither is cheese, but they are both vital ingredients.

How good would a pizza be without cheese, or worse dough?


And to this I have to say, I just don't believe you nor agree.

Firstly, you haven't defined relevance. What is it, and why don't you think Google has it, or has set "such a low bar"?

Next, 10% of a SERPs success relates to the algo chosen. The range of filetypes and poor coding a parser can handle, the database structure chosen, total size of the index (including how you get an OS to interface with so much memory), the number of factors that can be used in the algo, the time it takes to generate a results page, usability, refresh rate and a bazillion other factors all go into creating a "relevant" SERP.

To flippantly claim you could do better, with coin the only inhibitor, is not onlyunfounded and poorly justified, but downright ludicrous. At the very least, you would also need significant time to iron out the bugs.

IMHO, Google is incredibly relevant, especially when users are eloquent in what they want.

Even when they aren't, and buggered if I know what a searcher looking for mortgage really wants, Google offers:
1. A definition.
2. Spelling corrections for the dyslecix ;)
3. News on that topic
4. Ads from relevant vendors.
5. A calculator
6. Answers to Your Mortgage Questions.

Heck, I can even complain, which may be what I was after all along.

How relevant is that, and to how many people that typed in those words? I am extremely doubtful that there could possibly be a better, more relevant result to a wider range of people than that SERP.

'Course, I am more than willing to be shown how I am wrong.

The point is precisely that "relevance" is not something that can be readily defined in a short, off-the-top-of-the-head forum post. One of my main points was that when Google was founded, there was not in place either any really heavy-duty computational science regarding relevance, nor, by and large, the hardware firepower to implement it. Today there is, but it is neither simple nor, therefore, cheap.

One clear criticism of the present method is that it's so jejune that it can be almost completely summed in but three words: "links are votes". Nothing that naive can be terribly powerful. To say that a given task having not yet been attacked with currently adequate methods somehow proves that task undoable is a leap of logic that escapes my poor powers.

There was nothing flippant about it.

"If it hasn't been done, that proves it can't be done." And we're all still living in jungle trees and eating bananas. To announce ex cathedra what highly paid experts in a specialty field could or could not do is not onlyunfounded and poorly justified, but downright ludicrous.

You mean like Google--as we all very plainly see--still hasn't since the day they were founded?

--
Eric Walker
SEO Tools, Toys, and Packages

Disagreed. The algo can be summed up as "Links are a vote, not all votes are equal, what a page itself says is not the only way we ascertain what a page is about, relevant, important pages rank better for generic searches, the more information a surfer asks for from us, the more specific an answer we give, and the fewer percentage of results that will be totally relevant, and more pages indexed leads to better SERPs."

Now, that is one heck of a a lot more complicated than three words

You made some other comments that also found a bit odd, such as :

I just don't get what safe we have the combination to. Sure, we know PageRank as a comparison, but we don't know how it is used, or have any sort of granular accuracy as to the exact value, nor indeed a reliable or accurate tool to discover values, nor knowledge of when or how often this value is updated.

rather than a safe, it is like The Colonel's secret recipe. Sure, we all know there are 13 herbs and spices, but which 13, and how much of each.

Anyway, seeing as only you and me are left, where do you think Google is not relevant, and how do you define relevance? My definition is simple: in the absense of a semantically signifcant wuery, variety is relevance. Don't assme you know what i want, assume you don't, and let me choose the one I like best for myself.

Your definition?

I think people overlook the fact far too much that a link from a relevant page, ie one optomised for the same or very similar term, is worth far more than 100 links from sites that are not relevant. Even if the other relevant linking site is a low PR it is still helping to form a resource of sites and links that can provide the user with relevant information which is what they are trying to do in the first place!

Which would you rank higher a page optomised for a search term with 100,000 irrelevant links pointing to it or a page optimised for the same search term with 100 links from other sites optomised for the same search term? Which do you think would be more relevant??

It seems obvious to me that as people want to get the most relevant content then that is what search engines will try and link them with! This all feeds into the fact that a resource site or large directory / authority site will always do better then small content sites.

And how is a search engine supposed to filter these relevent pages.

will it search all the page that the link is comming from for the keywords.

normally links have discription text under them.

Wont directories be classed as authority sites because of the amount of similar themed links per section and relevant content per section?

Good post, but wasn't Gogole originally started in a Garage?

Dan

Yes (I think). That's part of its problem: it is a behemoth grown from an idea so naive it could be implemented by two guys in a garage.

Unless you are secretly GoogleGuy, you have no whatever of knowing what, if anything, past the original three words is so or is guess or is urban folklore. But don't take my word for it: look around every SEO forum you can find, and see what the veterans tell everyone SERPs success is all about: links. For that matter, do a Google right now on the term miserable failure.

I do not doubt that G has, over the years, done just what the makers of cars powered by the internal-combustion engine have done: pile on gadget after fix after doodad to try to first, overcome the severe inherent deficiencies of the underlying power plant itself, then second, to compensate for the further cockups introduced by the previous layer of patchwork. So now they have some contorted-in-six-directions variant of "links are votes", yet the fact remains nonetheless that links are votes.

Till they can bring themselves to wholly or largely jettison the ridiculous idea that what numbers of someones totally unconnected with the site will link to is what determines the site's value to a searcher, we will get the same endless stream of problems we get now--links, and the consequent SERPs, bought, sold, and traded like pig iron. I could show you right now a site, thriving, that offers ten thousand "relevant" links from a hundred IP-distinct domains with real content, for $6 a day, $180 a month, $2,160 a year, guaranteed--and credibly, from an honest and reputable company--to all be "legitimate" enough links that Google will take them. And the buyer sets the anchor text, exactly. Is this what searching is supposed to be about? The ability to accurately locate the sites that had the most money to spend?

And I think people overlook the fact there there is at present no way to establish relevance so easily--that is the whole point If they could so easily establish real relevance, this whole discussion would be moot.

Are vacation cruises "relevant" to incontinence? Are hip-hop groups "relevant" to skateboards? But people interested in marketing the first term of each pair will tell you exactly why advertising in places where people go to learn about the second term is a wise investment.

Any child can conjure instances of a reference on a page mostly about something else that is nevertheless, in context, relevant to something quite different. In a site page on scientific exploration in Antarctica, I suggest that one of scientists, out in the antarctic wind, asks another "Are we having fun yet?" (which is to be found in books of quotations), and I throw in a link on that phrase to the "Zippy the Pinhead" comic-strip home page.

Is a link from such a passing reference going to be downgraded because it didn't come from a page "relevant" to the target? Either people who think that it will are totally nuts or Google is totally nuts, take your pick.

The whole "links are votes" idea was, in essence, a sort of intellectual outsourcing: we can't figure out how to determine relevance, so we'll just let the cumulative mass of Toms, ****s, and Harrys decide for us, and if the ****s happen to predominate, hey, this is a business. It was, and is, a simple surrender: we can't figure this out, it's too complicated for us.

And so we get the web world we have today.

--
Eric Walker

Now you've really lost me. What is reloevance? Lets start there. I defined, in a detailed, example driven way what relevance is to me, yet you continue not to. Why? To claim you could write an SE that was more relevant than Google, then not define what they means, is unfair.

So what is worng with using links as part of an algorithm? I just don't get it. If you wanted me to build a house, I would want to be able to draw on every possible tool and raw material I could. I don't want anyone telling me that I can only use a hammer and steel, not timber.

That is what any discussion of PageRank ammounts to: a claim that the reduction in the available metrics that an algorithm "cheif" is allowed to utilise.

To then go off on this whole link buying argument is tired and innacurately portrayed.

Firstly, anchor text is not PageRank nor a vote. Its relevance to a discussion on PageRank being removed is tenuos at best.

Second, if you have a read of some of the historical complaints against PageRank, and WMW has HEAPS, it used to be how unfair it was to commercial sites because "..it is so much easier for non-commercial sites to get links". Now that link buying is a popular advertising option, the complaints have flip-flopped.

But the biggest issue I have with your post was that it started with :

and then continued to roll those very same myths out in defence of why links as votes are bad. In fact, the statement "what a page itself says is not the only way we ascertain what a page is about" is exactly what anchor text is, a feature you choose to attack two paragraphs later.

So, back to the start, I still don't know what you think relevance is, nor why using links to ascertain this is the wrong way to do things.

What does amaze me is that people seem to still be stuck on the idea that PR must simply be what it has always suspected that it is and that it has not evolved at all... that it is merely a measure of links and passing PR from relevant sites as they would be defined by similar keyword searches. I suspect that it is far more complicated than that, and while it is used as a component of a larger algo, it also has its own component elements and is an end all in itself.

The assumption is made that PR is strictly an inherited value from other sites, but we forget about the fact that G collects a great deal of unkown click information from cookies that could conceivably be utilized to create the foundation for an artificial search intelliegence. Isn't it possible that PR is a graphical representation of how tightly a page fits into a matrix of choices made by a market that makes other similar choices? If you typically search for and then click to visit a particular kind of site, you more than likely fit a loose profile of similar people with similar click habits. You in effect add to the matrix. Plugging you into the matrix and therefore identifying you as belonging to a similar market segment is easily done with the cookies on your computer.

With the ability to gather this information from millions of people, you can tighten this profile down significantly. It is simple statistical analysis. Amazon does it all the time. When you select one book, you are told that people that buy that book are also interested in several other books. Google can do the same thing. So when you go to Google and you search for specific terms and then select your clicks, it is reasonable that as long as your click fits within the established matrix of people that search by similar terms and make similar click choices, your click actually also contributes to PR.

If you and others who are similar start to select another choice, perhaps something offered to you from a new website, then by your clicks you start to add this one to the matrix as well. The more people click on it, the closer to center of the matrix that choice becomes, boosting the SERP relevance and the PR ever so slightly. If you think about it, this would also be why you sometimes see new sites and new pages get an initial bump near the top, then fall off for a time until they are appropriately placed according to their selected place in the search selection matrix for the profile of people that search by those given terms.

It would also be the basis for determining a wider field of relevance. So for instance if you are searching for exercise equipment and you also search for supplements, on the surface, they would seem to be irrelevant. But if thousands of people who search for exercise equipment also search for supplements, then the matrix starts to morph these two seemingly unsimilar searches into a profile. That information could also then be transferred over to determination of PR relevance so that if a supplement site links to an exercise equipment site, that link now adds to the relevance and more PR is transferred.

Okay... it's 2 AM here so if all that above doesn't make much sense, try reading it after midnight and it should make all the sense in the world