What's the problem in analysing googles algorithm?

Steve|munich

Spiderpig, Spiderpig!
Jun 13, 2009
114
2
0
Hi wicked folks,

let's assume google uses 200 factors to determine the ranking of a webpage.

what's the problem in making a "really good guess" of googles formula? Any math nerd laughs at a formula with 200 unknown factors when he has billions of data for use to make an approximation to the formula.
 


Do you have all of the data google uses to decide ranking?
Expected that question - no, not really everything, but quite a lot.

If you think about it, we have:
-- all onpage factors
-- most of the important links
-- data like CTR in the serps or bouncerate on the website
-- good guesses of things like search volume for the domain name
-- semantic data
 
Instead of analyzing google you should be analyzing the competition. They got to #1 somehow, right? Figure out what they did, and do it better. Tada.
 
Okay, so we can guesstimate 200 factors that Google uses.

Put those 200 factors, which the majority we aren't even sure are real, in a complex algorithm that took Google's team of genius engineers a decade to make.

Go for it buddy!
 
Hi wicked folks,

let's assume google uses 200 factors to determine the ranking of a webpage.

what's the problem in making a "really good guess" of googles formula? Any math nerd laughs at a formula with 200 unknown factors when he has billions of data for use to make an approximation to the formula.

Chances are it's not a formula (ax+by+cz=score) but a more complex algorithm. eg, in a given topic certain factors will be different from other topics.

As long as you've got the important factors, you don't really need to have all 200.

One approach would be to get yourself a few million pages and do your own calculations. Use a ML technique like SVM and you've basically got your answer. I know there are people doing this, I just don't know the results.

That said, you're going to get some form of answer that's along the lines of "get more content and links".

Sean
 
Any math nerd would tell you that if you have to many variables with no solid proof of any of them that the problem can't be solved.

hey can you tell me what the answer to this is:

something + something + something that might be + something that probably isn't but we aren't sure + something that was true yesterday but appears to be not today + something about mary + something is wrong in cleveland = ((xc)x(78)y-89823-2k2k2 (x)(y))i))
 
The difficulty in deciphering google's algo isn't the actual algo, it's the fuck ton of data they have that they use the algo on. Page Rank is a formula that analyzes a mesh of interlinking pages, and with the more pages interlinking, the more accurate it becomes.

So deciphering things in order to rank is way the fuck easier than deciphering things to determine rankings
 


Each participant was asked to rate more than 100 search ranking factors along with specific questions about hot issues in the SEO field. This document, representing the collective wisdom of expert practitioners, is, in opinion, one of the most useful resources for SEO practitioners of all varieties, helping to provide transparency into what matters (and doesn’t) for best practices in search engine optimization.​

The plural of anecdote is not data.
 
If you could figure out an algorithm to determine exactly how google ranks websites based on external data, you probably wouldn't be sitting here, making no money and dreaming about how one day you will crack googles magical code...
 
If you could figure out an algorithm to determine exactly how google ranks websites based on external data, you probably wouldn't be sitting here, making no money and dreaming about how one day you will crack googles magical code...

Even if you have all the details, it only helps you determine how to allocate your resources. Knowing the effect of inbound links on ranking doesn't help you beat a site with 10k inbound links if you've only got one link.

Put another way, knowing the secret ingredient to Coke won't help you topple them.

Sean