Lots of bloggers are commenting about this thread at Digitalpoint on why Google's algo has never been leaked. My take on the algo - it would barely help if it was public knowledge. Yes, we'd all sit down and analyze all the cool things Google can detect about trust and gaming and spam and manipulation... and then? Then, we'd go back to work trying to dream up good content and clever marketing plans to have our content receive links.
I'm not saying that there would be no room for clever spammers to build out some automated systems that might function for a few months or even a year, but they update that puppy pretty frequently. The only real value would be MSN, Yahoo! & Ask, who could use it as competitive intelligence. I guarantee that no matter how much you think you could make on the leaked Google data, those companies would find it far more lucrative.
It reminds me of a dialogue Rebecca, Matt & I had once joked about:
Matt: It's ours! We've stolen the precious Google algo from the 'plex.
Rebecca: Yeah, with our newfound knowledge, we can rule this puny Internet!
Rand: OK, wait - let's examine it and see what we need to do.
(six months of IR research later)
Rand: Aha! I've got it!
Matt & Rebecca (in chorus): What, what is it?
Rand: Links! We need links.
Last point of interest - if it ever was leaked... How would we know about it? Maybe it was leaked in October of 2004 and was the impetus for the infamous Florida update. Maybe the sandbox grew out of a leak. If you had it, would you really share with the whole world?
Honestly, at this point in the SEO game, it pays to be a great business strategist, a creative content developer and a phenomenal marketer far more than an analyzer of algorithms (and this is coming from the guy who spent nearly the whole of summer '05 reading nothing but patent apps and IR documents).
I remember reading somewhere that there are sections to the algo that each team is assigned to. They work on their piece and nothing else never seeing the whole code base. If what I read was true, only a few people have access to the full algo. I can't remember where I read it or if it's true, but if so that would do a lot to prevent it from getting out.
You guys are dead on again. There are so many moving parts it would take months to understand it. Unless you had a guide who knew the code in and out.
I'm surprised that nobody has suggested that the algo is so complicated & spread across multiple machines and infrastructures -- that even if it leaked, it would take a team of CIA cryptographers a week to make sense of it.
Rand just leaked the top-secret fact that Matt and I talk in unison all the time.
>>>Rand imagines in his head
Ok, I almost peed. Happy? ;-)
indubitably.
I think you guys at SEOmoz must know quite alot about the Google Algo, you build up links without even trying, I found this website because someone in a forum told me about your pageranking tool and said it was a really good way to measure search engine juice, good tool by the way, get that sandbox checker working as well because I think my site is in it, I want to know for sure though, my site is only 3 months old.
This is why I love this blog. Original content!
Oh, and diet coke here too Rand... and interesting theory on the Florida update.
Looking back at that update, and the timeliness of it (middle of Q4) made no sense for a group as smart as Google.
Here is the funny thing: having the knowledge of how it works doesn't automatically mean victory. I mean, it's common knowledge that Google's algo is largely a link-based algo. Now, assuming that links and domain age are in the top 10 factors, even if you knew the algo and figured out what the other factors are and what the weights are, that woudln't guarantee victory. If you try to rank on a super-competitive term where there are a bunch of 10 year old domains with tons of edu links and other high-quality links, knowing the algo only tells you one thing: that it's going to be extremely difficult to beat them.
At the end of the day the algorightm is only how the system works. Being successful in that system still requires real work. I mean, even if you had the recepie for KFC, you'd still have to make it and sell it to be successful and it would take years to be anywhere as successful as KFC is.
I agree with this post 100%. The algo is out... the answer is links, plus the basic knowledge of how to format a page - read any of the Guides to SEO and you will own the algo.
Carrying it out is the difficult part.
This is so true. There are no magic tricks to achieving anything worthwhile, and knowing Google's "secrets" would likely do nothing but confirm what everyone already knows. We know they don't like spam; we know they like quality links. Yahoo et al would have a lot more to gain than SEOs.
I realize this was probable more of a fun posting, but... :)
Let's just look at the ranking algorithm. Someone somewhere mentioned that they use over 200 signals, let's assume that's about right.
The algorithm could - if expressable in a single formula - be a function with those 200 signals, some weighting, perhaps some combined weighting with inverse signals... let's just simplify to an unrealistic: Rank(page.site) = weight1 * signal1 + weight2 * signal2 ... weight200 * signal200
The interesting parts would be the weights -- the main signals which are used are most likely known from publications and tests (right? at least simplified). So you'd be looking for the top weighted factors, things that could be "gamed". Suppose you end up with ... "links" -- what would that mean for your SEO strategy? No big change. Suppose you work on the "long tail of the algorithm", all those smaller factors that could add up: how do you play the system to target signals which are most likely only measureable when you have a body of content like Google has?
Now one thing that would be interesting would be to be able to play with the live algorithm for known sites: Get the 200 signals, their values for a known page/site, their weighting and see the output. Now tweak this setting and that signal - how does the output change? How far will a change in "link-value" take you? Could tweaking the lower 100 signals to the max make a difference in the niche or not? Which signals are inhibiting ranking? Which signals are maxed out? Which negative-signals (aka "penalties") have you collected and why? How could you reduce those signals while at the same time keeping the others up?
The algorithm as an algorithm written on paper is mostly worthless - what would be valuable would be direct access to the live internal structure of the system.
Here's the Algo:
Build an easy-to-use website that provides unique and valuable information.
Shhhhhh, don't tell anyone.
>>>Rand: Links! We need links.
Don't DO that. I almost got soda on my keyboard.
I'm always amazed at how many soda-drinking readers we've got here at SEOmoz... It must be part of the ethic of reading... Rand imagines in his head:
Rae: (is working away at computer)Rae: (thinks to herself.. hmmm, I should check out SEOmoz)Rae: (gets up to grab a soda first, not realizing that someone has implanted the suggestion in her brain that reading the site requires soda)Rae: (gets back, lifts up tab of can to satisfying aluminum tear sound)Rae: (clicks on SEOmoz, starts sipping, and follows this link Rand put up)Rae: (spits out a bit of soda onto keyboard, curses, throws it in the large bin on her left marked "all Rand's fault")Rae: (digs under desk for new keyboard and retrieves it)Rae: (forgives Rand eventually, realizing that he only had her best interests at heart)
Sure selling it to (insert other major SE here) might make you some dough, but that's nothing compared to....
Think about how many e-books you could sell for $197
!! Now, only $27 for the first 100--99--98 UPDATE only 13 copies left.
Plus these 301 other ebooks and software valued at... an incredible $5793, including lifetime master resell rights.
I still say that it has not leaked because Google does not have a dog named Duke.
are you sure about that TMS? There's a dog!
Roll that beautiful bean footage.