"TrustRank" vs. "PageRank"

[greeneagle]

TrustRank vs. PageRank

We have known for a long time that GOOGLE favors certain highly ranked (PR6+?) sites, bequeathing several advantages. They have also spoken of "Trusted Site Status". I have seen hints that “Trusted Site Status” is applied at various levels of benefit corresponding to certain trust levels by GOOGLE.

If you will notice though, this paper; "“Combating Web Spam With TrustRank”, YAHOO was instrumental in research and authorship.

I'd really like to discuss "TrustRank" how it is calculated, what it means, which SE is in the lead, how Google's "TrustRank" algs figure in, examples of YAHOO's "TrustRank" in SERP, how it interrelates to "PageRank" and whether it is really viably mature yet.

We always want to discuss “PageRank” but I think we have been missing an important and growing emphasis on “TrustRank”.

Ken

[greeneagle]

kgun,

It has occurred to me coming out off a 10 hour "hallucinogenic REM) sleep ("normal" coming down from a 44 hour workday) that we need to answer some basic linking questions in light of a potential "dawning" of "TrustRank" days ahead, if we can prophesy or speculate for a just a moment:

There is this old idea hanging around for a number of years that discourage OBLs, proclaiming that they "bleed off" PageRank. In fact we have alluded to it here, in this thread. Compounding that, from another direction, some of us have clients that are abjectly opposed to OBLs ---- period---!:

Let's put "PageRank" at "second seat", just behind "TrustRank" for a moment and try to answer a few linking questions from a "TrustRank" perspective, if we can, based on the math, but in plain English.

There are three (if I count right, correct me if I am wrong) statements made in the "TrustRank" algs that seem to be at the core of the calculations concerning linking and trust factor application for a site in the course of trust ranking review:

Three main factors TrustRank is based upon:

1) Good pages rarely link to bad ones. Bad pages often link to good ones in an attempt to improve hub scores.

2) The care with which people add links to a page is often inversely proportional to the number of links on the page.

3) Trust score is attenuated as it passes from site to site.”

These are core issues of the "PageRank" algorithms that seem to have already been being implemented to some degree.

It is MY OPINION that during the "Jagger" update that Google ran "smack up" against an "infrastructure" wall that extended that update beyond what they expected (6 weeks), resulting in a subsequent need to revise the existing infrastructure so that "TrustRank" algorithms could be implemented more efficiently. Now we are looking at a big bad beefed up "Big Daddy".

I'd like to hear your comments and opinions on that position and these linking issues from an expert and scholarly mathematical position, in plain English addressing those 3 issues.

<off topic issue>
No Televisions, Radios or any other background distraction... They corrupt working and resolving issues while sleeping.
<off topic issue>

Thanks,
Ken

[kgun]

one thing that is interesting is that “TrustRank” is directly tied into and dependent on the “PageRank” algorithms, making it easy to apply in the SERP.

Yes, if it is that simple.

[WebMasterKrames]

So I tried an experiment a while back, and it seemed to help for a while. I made words in my body like to web pages that ranked well and had info about whatever that word was. My Rank went up for a little while, but I don't know if it was from that (as always) but according to this theory, what I have done may actually be hurting me?
My theory was that these links may make my page appear more informative to the SE's. I'm not ranking so hot in Google currently (not like I used to before Jagger), but I show up nicely on yahoo and msn.

[greeneagle]

quoting kgun and paraphrasing? Albert:

"Make it simple, as simple as possible, but no simpler."

LOL

Did I fall off the stool?

Ken

[kgun]

Note the opening sentence of the 12 page pdf document:

"Web spam pages use various techniques to achieve higher-than-deserved rankings in a search enigine's results."
Source: http://www.vldb.org/conf/2004/RS15P3.PDF

Then, note the last post in the thread, what is spam:

kgun wrote:

SERP Spam:
By SERP spam we mean any thechnique, not related to content, that results in a manipulated position on the SERP's of one or more SE.

pemburung wrote:

SERP Spam:
pages on a SERP that are not relevant to the search term, have been manipulated to artificially raise their SERP rank over that which would be achieved normally through content, relevancy, and good design, or serve no useful purpose for a searcher. Also, the use of techniques, legitimate or otherwise, to achieve this.

DrTandem1 wrote:

Spam
is any repetitive use of a term for the sole purpose of artificially increasing SERP ranking.

This may end as a very informative thread if we could get some more constructive proposals. Then we may agree on a synthesis.

Do you see the relation?

And from the same abstract:

"We first select a small set of seed pages to be evaluated by an expert.

....

Our results show that we can effectively filter out spam from a significant fraction of the web, based on a good seed of less than 200 sites."

And note the last sentence in the conclusion:

"In a search engine, TrustRank can be used either separately to filter the index, or in combination with PageRank and other metrics to rank search results."

Important, interesting.

[kgun]

1. First of all, you can read a lot of the paper without any mathematical knowledge. Simply shut your eyes about mathematics and take what is written as facts with the given assumptions.
My: Assumption. They have not made any mathematical errors.

2. Simple matrix algebra.

3. Elementary statistics, simple probabilities.

You can ask (PM) me if you have problems with understanding a formula.

It is a very important paper, and should be read by anybody who want to understand how the pagerank and trustrank are computed.

[greeneagle]

About Google's "TrustRank" Trademark

Ken

[netndx]

good paper. thanks for sharing. I agree that this could help with the discussion on "what is spam"

[seolearner]

Originally Posted by kgun

You can ask (PM) me if you have problems with understanding a formula.

It is a very important paper, and should be read by anybody who want to understand how the pagerank and trustrank are computed.

I could not get through with the algorithm can you please explain it to me. As i want to understand how the pagerank and trustrank are computed.

[MtraX]

Nice find Ken. Now we can start with 'best practises to gain trustrank'...

[espectations]

In simple language I understand Trust rank as trust placed in informative sites that proofs that the site is owned by experts in that specific topic.

So, without taking the mathematics into calculation it is going to be all related to quality informative and educational content - right?

I am also of opinion from what I saw that if you write an article and on page link to (for example) a University publication, white paper or another article discussing the same topic, that page's trust rank will increase - it is kindof substantiating what you say with research.....

If the information is interesting and educational to the visitor they and others in the same industry will value the information and either link to it or as it so happened in one case they used my article to discuss the possibilities surrounding the topic in blog posts.

That provided a lot of trust in what the article was about so that page's rank increased substantially.

Good topic - thanks for raising it.

[kgun]

Originally Posted by seolearner

I could not get through with the algorithm can you please explain it to me. As i want to understand how the pagerank and trustrank are computed.

PageRank formula

starting with r(p) in section 2.2. PageRank.

You have a set of internet pages, let us say 20 billion.

Every page in this set has a pagerank, a number 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10

In essence the formula says, you find the pagerank of a page, let us call it P, by a weighted sum of the pagerank of all pages in the set. Pages that does not link to the page, our frind P, naturally get a weight of zero.

Then in greater detail, there is a decay factor, an outdegree (the number of outlinks from a page), and the number of pages in the formula. The pagerank of page q is in the numerator and the outdegree in the denominator. That means if two pages that link to page q have the same pagerank, but different number of outlinks, the page with less outlinks will influence the sum most (the number in the denominator is less). That page has a stronger vote for your page. Message, get links from pages with high pagerank and a low number of outlinks. The rest of the formulae is cosmetics. The decay factor makes it a weighted sum.

Before I continue, was this understandable?

If you do not know what a weighted sum is, here is an example:

USD amount | Weight
....100 ........... 0,2
....200 ........... 0,1
....100 ........... 0,5
....400 ........... 0,2

The weighted sum is:
0,2*100 + 0,1*200 + 0,5*100 + 0,2*400

Note that the weights in this case, as is often the case, sum to 1.

You may think, when you sum over all pages on the internet, you should get a number much greater than 10. Look at the wighted sum above, it can not be greater than the largest amount, USD 400, but that has a relatively low weight. The third amount carries half the weight and will drive the sum towards that number (100). So the highest number you can get is pagerank 10 for our page P.

If you have problems with the sum (greek capital sigma) notation, here are some good educational links:

Important figure, below the sentence

Notice that this notation has several elements labeled in the figure below.

http://www.math.montana.edu/frankw/c...igma/learn.htm

Note, j = index variable, 1 = start, j^2 = formulae and n = end.

This notation is not used in the paper, instead set notation is used. It is not different, if you sum over all index variables in the set, S e.g. Set = S = {1,2,3,4,5, ..., n}.

In essence, you get the first term in the sum by substituting the first value for the index variable in the formulae, then the next. You get the last term in the sum by substituting, in the example above n in the formulae. In the example above, you start by 1, then 2, 3, 4, .... n. Sigma notation for sum is in the end a simplified (compact) notation for a large sum. Think of writing out the sum for all URL's on the internet, and you see the point.

http://mathworld.wolfram.com/Sum.html

http://home.alltel.net/okrebs/page136.html

[incrediblehelp]

Remember "Trust Rank" could be nothing more than a human goggle editor, (Remember Google Eval last year?) adding this trust weight to major websites.

As we saw in some very generic keywords on Google the relevance is very exceptional. Why? I don't know, lets guess. One reason comes to mind that these websites have been given this "Trust Rank" factor SPECIFICALLY for that keyword queried. Remember one theory is when this Trust Rank is added to a website keyword is actually associated with it.

Interesting Slashdot on Trust Rrank as it applies to Google News.

http://slashdot.org/article.pl?sid=0...24206&from=rss

Now for PR, who cares, to me it is a little green bar.

[kgun]

Now for PR, who cares, to me it is a little green bar.

Yes, computed by one or another PageRank formulae :-)

There are a lot of formulaes, preferably with consistent rankings.

Should I introduce my own formulae that takes values

-10, -9, -8, .... ,-1, 0, 1, 2 ,3, ... ,8,9,10

-10 for very spammy sites :-)

Proposal:
Red bar for -10, -9, -8, .... ,-1.

Then we could talk about red and green lights in linking.

[greeneagle]

kgun,

It has occurred to me coming out off a 10 hour (near hallucinogenic REM) sleep ("normal" coming down from a 44 hour workday) that we need to answer some basic linking questions in light of a potential "dawning" of "TrustRank" days ahead, if we can prophesy or speculate for a just a moment:

There is this old idea hanging around for a number of years that discourage OBLs, proclaiming that they "bleed off" PageRank. In fact we have alluded to it here, in this thread. Compounding that, from another direction, some of us have clients that are abjectly opposed to OBLs ---- period---!:

Let's put "PageRank" at "second seat", just behind "TrustRank" for a moment and try to answer a few linking questions from a "TrustRank" perspective, if we can, based on the math, but in plain English.

There are three (if I count right, correct me if I am wrong) statements made in the "TrustRank" algs that seem to be at the core of the calculations concerning linking and trust factor application for a site in the course of trust ranking review:

Three main factors TrustRank is based upon:

1) Good pages rarely link to bad ones. Bad pages often link to good ones in an attempt to improve hub scores.

2) The care with which people add links to a page is often inversely proportional to the number of links on the page.

3) Trust score is attenuated as it passes from site to site.”

These are core issues of the "PageRank" algorithms that seem to have already been being implemented to some degree.

It is MY OPINION that during the "Jagger" update that Google ran "smack up" against an "infrastructure" wall that extended that update beyond what they expected (6 weeks), resulting in a subsequent need to revise the existing infrastructure so that "TrustRank" algorithms could be implemented more efficiently. Now we are looking at a big bad beefed up "Big Daddy".

I'd like to hear your comments and opinions on that position and these linking issues from an expert and scholarly mathematical position, in plain English addressing those 3 issues.

<off topic issue>
No Televisions, Radios or any other background perturbances... They corrupt working and resolving issues while sleeping.
</off topic issue>

Thanks,
Ken

[kgun]

Originally Posted by greeneagle

Three main factors TrustRank is based upon. I would add, as TrustRank is defined / computed in that paper:

1) Good pages rarely link to bad ones. Bad pages often link to good ones in an attempt to improve hub scores.

Yes.

2) The care with which people add links to a page is often inversely proportional to the number of links on the page.

Yes.

3) Trust score is attenuated as it passes from site to site.

Yes. Note the following sentence on page 580.

"These obervations suggest that we reduce trust as we move further away from the good seed pages".

I regard the initial seed pages as a filter that get thinner and thinner the further away you move from the initial seed.

Note: Page 577 of the paper:

1. It is one method among many possible methods.

"The methods that we present in this paper could be used in two ways: (1) either as helpers in an initial screening process, suggesting pages that should be examined more closely by an expert, or (2) as a counter-bias to be applied when results are ranked, in order to discount possible boosts achived by spam".

2. It is subjective, since it requires human intervantion.

"Since the algorithmic indentification of spam is very difficult, our schemes do not operate entirely without human assistance".

[williamc]

Very good debate going here guys, on both sides. However, for the newbies reading this lets point out a few simple things please.

#1. TrustRank has not been evidenced to be in use.
#2. Matt Cutts himself stated that intelligent people wont read too much into the patents.

This thread is all about speculation and should not be taken to heart until trustrank has a whole lot more evidence to back it up.

[kgun]

This thread is all about speculation and should not be taken to heart until trustrank has a whole lot more evidence to back it up.

As it all about speculation to assume that it is not used in one form or another.

Fact:
The paper discusses one method among many, formalizes trust rank and indicates how it can complement PageRank. Figure 10 page 585 speaks for itself.

Conclusion of the paper:
TrustRank can be used to fight / reduce web spam.

[williamc]

Originally Posted by kgun

As it all about speculation to assume that it is not used in one form or another.

Agreed on both this and the fact that it could be used to fight spam.

[swirt]

I think the Trustrank approach is interesting, not so much because it would influence the serps and make them more meaningful on first pass. I think where its real value lies is in driving people to make better sites with better content. This is something that pagerank based on links has never driven. Pagerank practically encourages spam, while Trustrank could encourage better content and thoughtful linking.

The question is, is there evidence that it is already in play?

[Mel]

Lets review a few things should we?

First off this is a paper written primarily by Yahoo experts (the Stanford Professor is a consultant to Yahoo) and it would be rather strange for Google to base a ranking algo based on Yahoos work.

Second the registration of a trademark indicates an intent to use something called Trustrank, but that does not mean its the same algo as written up in the Trustrank paper by Yahoo.

If we do decide that the algo is the same as Yahoos, then it has to have human intervention lots of it and that is normally not the way that Google works.

Is there any sliver of information that indicates that this discussion is based on anything substantial? Or is it based only on a similarity in names?

[greeneagle]

Start here Mel:
http://blog.searchenginewatch.com/blog/050118-204728

If Google sees nofollow as part of a link, it will:

NOT follow through to that page.
NOT count the link in calculating PageRank link popularity scores.
NOT count the anchor text in determining what terms the page being linked to is relevant for.

IMO - It's not only a "bloggin deal" anymore.

Ken

[swirt]

Originally Posted by greeneagle

NOT follow through to that page.
NOT count the link in calculating PageRank link popularity scores.
NOT count the anchor text in determining what terms the page being linked to is relevant for.

If only it would say that it would NOT count the outgoing link against the site in any way, then there might be a little light shining into the black box. ;-) But essentially they are saying that, so it is interesting that you tie the nofollow together with the trustrank theory. I guess the telling question would be, of the sites that improved after adding the nofollow, are there any of them where the rise can not be attributed to reducing PR bleed? (I am having trouble imagining a case that would allow us to distinguish between the two possible causes.)

If the nofollow is applied to all posted outgoing links, then from a trust point of view, the site does no wrong by pointing to bad sites, but it does no right in pointing to good sites either.

[greeneagle]

Recently the TrustRank vs PageRank issue has been resurfacing in some threads.

A while back we started this thread, before "TrustRank" influence may have been more directly implicated in alg shifts.

It appears that "TrustRank" is a module that is designed to easily "plug and play" with "PageRank" and that we have seen some implementation in recent GOOG alg shifts. I am seeing some correlation there especially where "TrustRank" seems to better define "healthy" linking practices.

I believe continuance of this thread may have more "relevance" starting with "Jagger" and extending not only through "BigDaddy" but into the future, from here.

Ken

[scanmonkey]

Originally Posted by incrediblehelp

Remember "Trust Rank" could be nothing more than a human goggle editor, (Remember Google Eval last year?) adding this trust weight to major websites.

As we saw in some very generic keywords on Google the relevance is very exceptional. Why? I don't know, lets guess. One reason comes to mind that these websites have been given this "Trust Rank" factor SPECIFICALLY for that keyword queried. Remember one theory is when this Trust Rank is added to a website keyword is actually associated with it.

Interesting Slashdot on Trust Rrank as it applies to Google News.

http://slashdot.org/article.pl?sid=0...24206&from=rss

Now for PR, who cares, to me it is a little green bar.

Trust rank is like a "get out of jail free card"

The idea that "a human goggle editor giving sites Trust Rank for a keyword that is actually associated with it", makes sense to me.

That way the serps at googl* wouldn't get too monkied up on aglo changes and adjustments.

[kgun]

Read this related thread about pagerank

Questions:

1. There are various fund ratings where I personally prefer the Standard & Poor rating. Will PageRank / TrustRank be a measure analogous to fund ratings in the future?

2. Which SE will be most trusted in the future? Can a little green indicator be more important than anybody can even think of today? It has been said by experts, that the worst dictator you ever meet is "Mr Market." Is the little green indicator the potential for a new dictator or is it only gimmic?

3. Directories versus search engines. As web spam increases, will directories become more important?

4. Will there be a "Standard & Poor SE" and a "Standard & Poor Directory"?

5. How deep is the measure of trust? Will economic theory like "the theory of assymmetric information" and "game theory" be used to combat web spam.

6. Subject for a Doctor thesis.

The race between web spam and trust.

[Tubby]

Kgun asked
3. Directories versus search engines. As web spam increases, will directories become more important?

I think so.

[scanmonkey]

Originally Posted by kgun

3. Directories versus search engines. As web spam increases, will directories become more important?

There's a lot of good directories out there. The problem is that most users don't know they exist.

[BanquetTables.Pro]

Originally Posted by greeneagle

Start here Mel:


If Google sees nofollow as part of a link, it will:

NOT follow through to that page.

Ken

BS it will follow the link, but not pass juice on.

[williamc]

Banquet: welcome to 2006

[Feddy]

i am aware with Page Rank but not trust rank but from here i understand about it so thanks to WPW forum and all.