June 2010 Tag Rankings

[ipopbb]

For those who have been following my threads...

(This is really interesting to me because its my first "post Google MayDay Update" look at the data. I find it quite shocking AND it parallels some of what I think I am seeing in my keyword space. )

It also supports a new theory I have regarding link juice... its called:

"You HAVE to link to drink."

Here is the data:

Tag Rankings for June 02, 2010

java decodeTags -h -d 20100514.txt 20100602.txt 
RANK	CHANGE		TEST
1	u + 4		linking to relevant pages
2	new   		dl > dt > dd
3	u + 20		body
4	u + 8		table > caption
5	u + 5		textarea
6	d - 5		title
7	u + 27		h3
8	u + 1		pre
9	u + 8		sub
10	d - 3		cite
11	d - 8		acronym
12	d - 8		s
13	d - 11		samp
14	new   		(meta-description).value
15	u + 25		(meta-keywords).value
16	new   		table > tfoot > tr > td
17	u + 8		table > tr > td
18	u + 12		del
19	new   		q
20	u + 2		sup
21	new   		h4
22	d - 16		h5
23	u + 10		h1
24	u + 8		dfn
25	d - 14		kbd
26	u + 11		select > option
27	u + 11		(input-input).value
28	d - 13		bdo
29	u + 6		img.alt
30	d - 11		u
31	u + 13		html comment
32	d - 16		strong
33	d - 12		strike
34	u + 7		(meta-author).value
35	d - 7		i
36	d - 23		small
37	d - 19		abbr
38	u + 1		a.href
39	d - 13		dl > dt
40	new   		font
41	d - 17		tt
42	new   		big
43	      		b
44	d - 13		ins
--	was 42		document
--	was 20		ul > li
--	was 27		em
--	was 36		a
--	was 14		code
--	was 29		caption
--	was 8		p

 

[J0kerz]

H1 tag ranking better than H3? Huh?

 

[ipopbb]

H1 has consistently been a poor performer for over a year in my data. What that means... who knows... but you can read my other threads if you want some history on the matter.

This data is just that... data. Google sorts my test pages that each use an imaginary word in a single tag or attribute test test. I search for the keyword in Google and google sorts the test cases. This data is the movers and shakers in Google's sorting of the tags tests since my last measure.

I'm not recommending or suggesting anything.

Just sharing my data.

 

[ipopbb]

Oh yeah...

HOLY SHIT! HTML comments went way up! EXPLOIT! EXPLOIT! EXPLOIT!

 

[goawayplease]

"Link to Drink" is kind of old news... in GoogleSpeak it's called being a good neighbor. Highly Authoritative content will often link to relevant content.

Seeing some quantitative data is always nice though. I'm liking the way you think.

 

[scb335]

Really appreciate your repeated sharing of these findings. It's interesting and educational to watch the movements over time. Keep up the great work!

 

[ipopbb]

"Link to Drink" is kind of old news... in GoogleSpeak it's called being a good neighbor. Highly Authoritative content will often link to relevant content.

Seeing some quantitative data is always nice though. I'm liking the way you think.

True... in my imagination I'm seeing a slight variation to it now... the "YOU HAVE TO" part... Something to the effect of the amount of outbound linking you have is proportional to the upper limit of link juice a page can receive... the subtle variation is that linking changes the size of the cup a page can use to hold an amount of link juice. Or so the hypothesis goes...

before it was an additive factor... you weren't penalized for not doing it. My site splash pages rank wonderfully and the link everywhere. My leaf node product pages used to rank wonderfully and now not so much... they don't link hardly at all. I think linking might play a role in the drop off of long tail traffic to those pages. I plan to deploy tests for this hypothesis later this week. I'll share what I find of course.

 

[MisterGemini]

Oh yeah...

HOLY SHIT! HTML comments went way up! EXPLOIT! EXPLOIT! EXPLOIT!

Does that mean this has changed from your original post? They were originally 31.. they have gone up more?

 

[MisterGemini]

True... in my imagination I'm seeing a slight variation to it now... the "YOU HAVE TO" part... Something to the effect of the amount of outbound linking you have is proportional to the upper limit of link juice a page can receive... the subtle variation is that linking changes the size of the cup a page can use to hold an amount of link juice. Or so the hypothesis goes...

before it was an additive factor... you weren't penalized for not doing it. My site splash pages rank wonderfully and the link everywhere. My leaf node product pages used to rank wonderfully and now not so much... they don't link hardly at all. I think linking might play a role in the drop off of long tail traffic to those pages. I plan to deploy tests for this hypothesis later this week. I'll share what I find of course.

So you are saying there is more of a ratio for link building capacity based on the links given. Example my site links out to say 10 websites, then maybe 100 links will count as juice towards my site (assuming a 1:10 ratio).

You think the ratio is something different?

 

[snwbrdstylee]

ipoppbb,

Thanks so much for all of your informative posts, they have opened a door that has me thinking outside of the box and diving more of the core quantitative SEO issues, which is something I have been lacking for quite awhile.

Any chance you would think of starting an email list to share your thoughts with hungry minds?

 

[ipopbb]

No. It means I glossed over its significance in the OP. New all time high.

Does that mean this has changed from your original post? They were originally 31.. they have gone up more?

 

[ipopbb]

So you are saying there is more of a ratio for link building capacity based on the links given. Example my site links out to say 10 websites, then maybe 100 links will count as juice towards my site (assuming a 1:10 ratio).

You think the ratio is something different?

Or think of it this way... if you don't link out enough, you might be spilling link juice on the floor because your cup isn't big enough. I find this notion sensible because it reminds me of fluid dynamics... determining flow and pressure limits as pipes branch into ever smaller pipes then drain into sewers.

There is only so much water that can flow through your page. You need a conceptually fatter pipe to handle the flow. More linking and faster page load times might be how google judges this... but its all hypothesis.

 

[ipopbb]

That would be like spamming spammers. Unthinkable!
Nah... I'll just post here.

ipoppbb,

Thanks so much for all of your informative posts, they have opened a door that has me thinking outside of the box and diving more of the core quantitative SEO issues, which is something I have been lacking for quite awhile.

Any chance you would think of starting an email list to share your thoughts with hungry minds?

 

[ultimategamechain]

This is good information. Never really thought about it this way.

 

[ipopbb]

I know we differ in opinions about this.
But wouldn't it be more valuable information to extract the number of characters residing between the tags?

Maybe...Probably... what is "valuable" to one person might not be to another. I have made such measurements in my other thread "More Data from the SEO lab". I found it quite valuable.

For this experiment however:

In most cases there is 1 word in the visible content and its the same word from test to test. The data shows level of nesting and they all reside in the same generic <html><head></head><body></body></html> template.

Being more or less nested doesn't show any pattern in the data, especially if you look historically.

I suppose I could double the test cases and have a version with two keyword matches in every case, but I already believe in keyword density and google's sweet spot for document size. I know from my thread "More Data from the SEO lab" that 4 title matches is no more powerful than 1 title match in placing on the first page. And 5 or more matches is a one way ticket out of the top 100... This is contrary to the behavior of keyword density. That's because page titles are treated as an exception to the fundamental behavior and have their own set of rules...

I am looking for entities with special behaviors. The fundamental math of indexing texts for keywords gets you into the top 100 for your keywords. It doesn't explain the sort order, at least not for any of my keywords. Its my belief that it is all the various artificial influencers that determine the sort order. So, I'm looking for patterns over time.

I can't explain why my data behave's this way. I'm reluctant to stop watching and sharing it until there is a plausible and practical understanding... or until I burn out and do something else... The fact that for over a year now h1 has stayed low on the list and title has stayed high on the list hints that I'm looking in the right direction. I think I understand title. I am very curious still about the rest.

I have provided my test case source. So please feel free to use it any way you like. I can't do everybody else's experiments for them. I barely have time to do my own. However, I will continue to share my data. I hope others begin to do the same.
M.A.D., honestly and genuinely, I think you are smarter than I. I want nothing more than access to YOUR data. hint hint! I think we all can benefit greatly by working openly and sharing our findings and theories. I have "a" perspective. You have another. Sounds like a match made in heaven. Share some data and lets brainstorm on it. I bet you could blow my data out of the water if you were motivated to do so. So, "wow" us. Don't rush. It may take weeks and months... might take you minutes and seconds... As much as others like pouring through my data I sure wish I had more to work with.

Respectfully,

Ted

 

[niksanyl]

Ipopbb,
I've been following your threads with great interest. I hope others realize what you've provided here and take some action. (not my competitors of course). If nothing else, do some testing. Which I will be doing soon, and would be happy to share my data.

I have a quick question. I have an ecommerce store with 35K products. Should I ignore the page rank leak from having 35K outbound links,and actually link to each of the manufacturers websites from each product page? I've done some linking to places like wikipedia from some of my major keyword pages. I'm just a little nervous about linking out to widgetmanufacturer.com from each product page when i sell widgets. Obviously, I'm trying to build pages that are the best for my customers, not google, but it's hard to ignore the math. Your thoughts?

Niksa

 

[niksanyl]

Also, I don't see blockquote in your data above. Last month, it was pretty high. How did it shake out this month?

 

[unzero]

I will follow these threads, so valueable information!

 

[naviivan]

Since you use a unique keyword, what is the page you are linking out to? I mean it says "linking to relevant pages", so what would a relevant page be for a unique gibberish keyword? Or is it just linking out to any page?

Does the outbound link have the rel=nofollow tag? Have you tested it with that and are the results different?

 

[ipopbb]

Since you use a unique keyword, what is the page you are linking out to? I mean it says "linking to relevant pages", so what would a relevant page be for a unique gibberish keyword? Or is it just linking out to any page?

Does the outbound link have the rel=nofollow tag? Have you tested it with that and are the results different?

My source code is available in another thread if you'd like.

A. the tests are named 001.html 002.html etc...
B. the index links to the tests without ever mentioning the secret word anywhere in the source or URL.
C. the index hit #1 in the sort order... It amazes me too.

By relevant I mean pages that are tuned well for the term. They can be internal or external to the domain.

I have not tested rel=nofollow with anchor tags.