Forum Moderators: open

Message Too Old, No Replies

SERIOUS Google update algo analysis thread (Dominic)

NO whining or cheering about how your site is doing in this one.

         

rfgdxm1

6:21 am on May 6, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



This is a continuation of an idea for a thread that I started a few updates back. The topicality is listed below, and the expectation is that this thread will be restricted to just that. In the case of this Dominic update, GG has stated that other aspects of the update will be rolled out as the update develops. Thus, for this update it is possible that the observations made early on will not hold true by the end of the update. This is OK, because if patterns like this hold true for later updates, members here can use the search feature to find this thread and see how past updates developed.

----

I'm starting this thread because another member suggested such would be a good idea because the main Google update thread is cluttered with posts like "OMG, I've been dropped in the new index!" and "Yippee, I'm now #1 on a key SERP". This thread is ONLY for serious, generic discussion of changes that you are observing with the new algo in this update. As in things like "Looks to me like PR is less important this month, and anchor text of inbound links counts more.", etc. How your site is doing has no relevance here unless you can explain why you think so in terms of a general algo update.

HayMeadows

4:31 pm on May 6, 2003 (gmt 0)

10+ Year Member



"Google's order of results is automatically determined by more than 100 factors, including our PageRank algorithm."

Could it be that we are seeing one factor applied at a time starting at factor number 50? Whereas before we saw factor 102 and 103 applied only?

I have a hard time believing that Google has asked us to play by the rules and will now change everything. I'm sure if this is the case - there will be a lot more people inclined not to play by the rules.

And they know that....so we must have a long way to go before we see the final results.

seasalt

4:36 pm on May 6, 2003 (gmt 0)

10+ Year Member



Allergic:

>>"Adaptive Methods for the Computation of PageRank" - Dominic based in part with this new algo?<<

After reading the abstract from the link you gave, I would say in part, yes, this is what is going on. Google starts with displaying a basic index(what we see on SJ, etc.) since most "pages converge to their true PageRank quickly". Then through a series of further iterations, PageRank on the remaining pages is computed. The "PageRank of pages that have converged are not recomputed at each iteration after convergence." Saves Google computation time. Maybe time for more frequent updates in the future.

Since "slow-converging pages are generally those pages with high PageRank", I would guess that is why so many SEOs on WW are noticing drastic changes. We just have to wait for the further iterations. I think it will all balance out for the better in the end.

adsoft13

4:47 pm on May 6, 2003 (gmt 0)

10+ Year Member



Hi all,

Sorry if I have mentioned something you already know ...
But to me this is revolution in google algo :)

I have web site, which is targeted to "kw1 kw2".
The site appeared to be #1 for it. Noew it is #7 for it. I suppose because of lack of incoming links ...
BUT
only 10 mins ago I have got that it is #1 for "kw1 kw3".
And I didn't have kw3 not in title, nor in description and NEVER on my whole site at all (it is just 3 pages ;) ). And I know I ODN'T have it in any incoming link. And previously it couldn't be find at all by this combination.
In fact this combination is more competitieve then "kw1 kw2".

It is something, which caqn not happen at all with previous google algo. And this is revolution.
I suppose now google uses not just direct words database, but they combine them by semantic, and tries to find similar words and combine results, which can not even have anything in common with what you EXACTLY search for, but can have the same theme by the words they use. F.E. you search for computers and the first results are PCs ...

Great - isn't it? :)

c1bernaught

5:05 pm on May 6, 2003 (gmt 0)

10+ Year Member




I'm noticing that sites that entered the index last month are not showing up in the serps...... yet.

Looks like all backlinks are accounted for.

Still early though and this is a new game.

Exciting.... right?

dupac

5:07 pm on May 6, 2003 (gmt 0)

10+ Year Member



I seen my site going down inthe new results but seeing the links going in are really less as compared to last month.
I feel that google is not taking intoconsideration all the incoming links, its kinda old database or they are not counting certain link coming to site. which maybe the links comin to page from my own site.
deepak

CCowboy

5:22 pm on May 6, 2003 (gmt 0)

10+ Year Member



My older sites have stayed In their top positions. My new sites have now beaten out the former top result holders for each of the new site's keywords. They have only beaten them out by 1 or 2 places.

It seems to me that links from inbound anchor text of site links with higher PR counts more. My keywords are from anchor text from pages of PR5, PR6 and PR7 with multiple links from each site.

The sites I rose above only have inbound links from PR4, PR5 and PR6 sites and have fewer anchor text links from each of those sites.

HenryUK

5:47 pm on May 6, 2003 (gmt 0)

10+ Year Member



New results? I don't think we can see the new results.

for kw1 kw2 I am getting the following different results:

-fi: #8 of 5,080,000 (no backlinks)
-sj: >#100 of 4,480,000 (no backlinks)
-zu: #5 of 3,960,000 (100+ backlinks)
www: #6 of 3,920,000 (100+ backlinks nb same figure)

www2 and www3 seem identical to each other and to -sj.

That trio also show mirror sites for #1 and #2, so I guess there are spam filters yet to be applied.

It's -fi that puzzles me the most - it seems to be taking account of links in (by giving me the ranking), but not actually showing them.

My best guess is that www is still the old index, -zu is some kind of everflux version of www (!?); -sj is the first version of the new index and -fi is a *partially* despammed and PR-reranked version of -sj with more actual backlink data to be brought in.

However - for a range of other kw combinations (relating to content pages rather than the home page) the actual rankings of individual pages don't alter much. Apart from -sj where my own former inadvertent "spam" (duplicate domains indexing twice, long since 301'ed) have come back from the dead.

NB - no complaints, just fascinated about what is going on.

PatrickDeese

5:56 pm on May 6, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



I have a site that is top 15 for a highly competitive search phrase (4 million + results) and in SJ it has dropped about 30 positions.

Since this site is all graphics on the front page, and relies solely on ALT tags, title tags, supporting "real text" content on interior pages and anchor text in the incoming links, I would say that the weight of ALT tags has diminished considerably.

This may be a result of "disappearing" backlinks (from 134 to 54), but it is a little disconcerting to see a dramatic drop.

I can only wait and see what happens when the new index is live and using current results.

Oaf357

6:07 pm on May 6, 2003 (gmt 0)

10+ Year Member



I think it's a little premature to analyze the algo when the update hasn't even gone live on WWW yet (from what I can tell).

bnc929

6:08 pm on May 6, 2003 (gmt 0)

10+ Year Member



Observations:

1. I have a page that is #1 for it's very competitive search terms. It has a PR of 6, and has only one incoming link (an internal one).

This page is still #1 on every server I've checked.

2. I bought a new domain a couple months ago and started to develop it when I realized it was banned. I emailed Google a couple weeks ago and they said they'd look into it. In the new indexes it is no longer banned.

french tourist

6:11 pm on May 6, 2003 (gmt 0)

10+ Year Member



for kw1 kw2 I am getting the following different results:
-fi: #8 of 5,080,000 (no backlinks)
-sj: >#100 of 4,480,000 (no backlinks)

NB - no complaints, just fascinated about what is going on.

I am fascinated too... now -fi knows 187 pages on my website, while -sj knows 185. I tried to find out which were the two newly discovered "by hand" but it did not work : serps on "site:www.mysite.com -dqfdqf" began exactly the same on -fi and -sj but diverged after #70!

(There remains a long way before I recover the 5630 pages known by regular Google, but I keep hope -))

(Edited half an hour later : I have checked out with a script which were the two pages known by -fi and not by -sj : nothing to understand, they are seemingly random pages from the middle of the website, not Freshboted recently, not pages receiving links, the real "average" pages)

[edited by: french_tourist at 6:45 pm (utc) on May 6, 2003]

adsoft13

6:26 pm on May 6, 2003 (gmt 0)

10+ Year Member



HenryUK,

Great finding ... -fi seems to have -sj algo but with some more backlinks and better results ... at least they seems to be less spammy and more realistick (the good sites in my industry returns to top positions on this index).

needinfo

6:41 pm on May 6, 2003 (gmt 0)

10+ Year Member



adsoft13
I would have to disagree with you there. Many of my searches on the fi server for my industry are returning sites specific to "blue widgets" or "red widgets" when the search was for "green widgets" for example. Comparing this to the current www results it's far inferior. However sites do seem to be less spammy for sure.
I seriously can't imagine Google releasing search results such as the ones I've seen. This would be returning in some instances nonsensical results to the end user...something Google does NOT want.

I think there is a while to go yet. My theory is that in the past we've seen the last few steps of the update when it showed on WWW2 & 3, whereas now we are seeing the majority of it.

P.S. My sites are doing ok on the fi server.

[edited by: needinfo at 6:43 pm (utc) on May 6, 2003]

Allergic

6:43 pm on May 6, 2003 (gmt 0)

10+ Year Member



I agree to Adsoft13 for -fi. -fi have the biggest index by breaking the 3,5 billion barrier with 3,71 billions occurence of the word "the".

Zapatista

7:30 pm on May 6, 2003 (gmt 0)



Sorry, haven't had time to go through all these messages but I have a theory that older sites in this algo are receiving a boost. I know of one, registered since 1997, has terrible PageRank, ranking above higher PR sites. That's the only reason I see it could be doing so well.

HenryUK

9:36 pm on May 6, 2003 (gmt 0)

10+ Year Member



thanks adsoft

However, I do agree with needinfo about the overall quality of -fi results.

*My own spam* is showing in these results - by which I mean old pages which were double-indexed when I had two domains pointing to the same site. After advice here and a temporary ban I managed to get the relevant 301s organised and my "spam" results disappeared in the last update.

(it was all due to ignorance and not *intentional* spam, all cleaned up as soon as I worked out what I was doing wrong)

I think there are more filters to be applied to the results on -fi; where that "iteration" will appear is anyone's guess!

onebaldguy

10:02 pm on May 6, 2003 (gmt 0)

10+ Year Member



fi looks a little cleaner than sj.In my industry, sj is showing two sites with identical content, both ranking in the top 5. fi only shows one of them - which makes it cleaner.

However, there are still two separate companies in the top 5 which are simply part of a big link network. They both have about 80 pages, and they link all of these pages to each other. Those are the only backlinks showing (with the exception of a Yahoo! link). I think most people in the forum have several sites and will include some links to a few of their other sites and I would not consider that spam and I do not think google would consider that spam. But it would be hard to understand how interlinking this many sites would not be detected by google (neither Teoma nor alltheweb have either of these sites in the top 10.)

It is possible Google is slowly applying certain filters. One that would detect heavy interlinking would probably be implemented only after a higher # of iterations. So maybe this filter has not been applied yet.

HenryUK

10:08 pm on May 6, 2003 (gmt 0)

10+ Year Member



Yep, -fi is better than -sj but still hoping that -final is better than -fi...

;-)

badger_uk

10:16 pm on May 6, 2003 (gmt 0)

10+ Year Member



Maybe google are applying certain filters to one database then moving the new results into another database. Just a thought.

badgeruk

RawAlex

10:17 pm on May 6, 2003 (gmt 0)

10+ Year Member



One site in -fi is showing my page as 9k, while -sj is showing 10k, with the newer version. This is a 3 week old change, so I am not sure that -fi has the latest data in it.

Alex

albert

10:49 pm on May 6, 2003 (gmt 0)

10+ Year Member



My 2 c for -fi and -sj:

-fi shows more of my old and new pages, but not all of them (compared to old index), and is generally showing more sites in results.

-sj shows less than -fi.

-fi and -sj are showing same amount of backlinks which is less than old index.

-fi shows less but different 'spammy' sites than -sj.

All may be strongly related to my market / keywords aso.

ariff44

11:08 pm on May 6, 2003 (gmt 0)

10+ Year Member



I agree: -fi certainly seems more accurate than -sj.

So the question is, which has more current results? Does it go

-sj
-fi
and then www

or

-fi
-sj
and then www

albert

11:12 pm on May 6, 2003 (gmt 0)

10+ Year Member



Who knows ... but:

-sj started to show this update stuff.

Given that together with GoogleGuy's posts, and my optimistic nature:

-sj > -fi >? > www

[edit]forgot to say: "?" another interesting part of that update[/edit]

annej

3:34 am on May 7, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



An interesting observation, a page I just put up a couple of days ago is showing in serps with a freshdate of 'May 6, 2003' (today). Freshie must still be active in the midst of all this. I was just checking on plain old www.

Hmm, It shows in -fi and everything else I tested as well. Sometimes it shows the fresh date and sometimes not. This is a brand new page, it wasn't around till a couple of days ago.

mrguy

3:56 am on May 7, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



For the words I'm watching, I am now noticing a difference across all 8 datacenters.

What's funny is, none of them reflect what is on the SJ server.

Perhaps the filters and backlinks are being added in as they propagate across the other centers.

c1bernaught

5:58 pm on May 7, 2003 (gmt 0)

10+ Year Member



I would say that data is being added back in at this point. A domain that was missing just reappeared on both sj & fi. No listing in the serps on either though....

twilight47

6:09 pm on May 7, 2003 (gmt 0)

10+ Year Member



"GoogleGuy: I'm expecting SJ results to show up at other data centers, then gradually over time we're going to pull in newer spam filters, backlinks, etc. "

GG has said this basic comment in different threads. My question is define "over time"? Is that days, weeks, months?

Mohamed_E

6:17 pm on May 7, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



Yesterday morning about 30% of my pages were missing on -sj. Late last night only about 20% were misssing, so they seem to be getting them back at a reasonable clip.

PollyG

6:25 pm on May 7, 2003 (gmt 0)

10+ Year Member



Hi,

About 20% of my pages have also re-appeared over the last day on -sj and -fi.

I'm hoping they will stick as my SERPS have improved dramatically - thanks to all the tips on this forum hopefully!

maardsma

6:59 pm on May 7, 2003 (gmt 0)

10+ Year Member



Hi All,

I just noticed that a new backlink that I received only a couple of days ago is showing up in www when I do a "link: mydomain.com" search. I don't know enough about Freshbot to know if the site that added my link is indexed by Freshbot or the regular bot, but the listing in Google does have a cache entry and the cached content is only a couple of days old. The date Google shows next to the entry is May 6 (yesterday).

The backlink doesn't show up in -sj, www2, or www3.

Does this mean www is showing non-Freshbot updated pages from only a couple of days ago?

Mark

This 263 message thread spans 9 pages: 263