Forum Moderators: open
SEO and Search Engine Optimization are used throughout the page in equal proportion.
We are #2 for "SEO state" post Florida while "Search Engine Optimization state" has dropped out of site.
Seems odd that we a good page for one and not the other.
Thoughts?
Eventually, Google will realise this is stupid. As I wrote in a new thread that never appeared, pages do not change between being spam and non-spam millions of times a day depending on user searches.
Eventually, either Google will drop this ludicrous concept or users will drop Google.
Kaled.
I have a 3 word combination that I was #1 for before this update. Now it's all the directory sites and links pages, etc. But for the two word combinations of the same three keywords I'm on page one.
From what I suspect about the new algo, it will just be a matter of time before the serps clear up for three keywords, then four, then five... Just be patient for another week or so, things should smooth out again.
I think you are on to something with your more than 2 word filter.
One of my clients sites was consistantly #1 for "los angeles widgets". Since Florida is no where to be found.
However if you search for "angles widgets" guess what, right back on top.
However, I have discovered that searching on the singular form of the searches usually produces the old pre-Florida results or similar to them.
The plurals (and more important terms on my index page and main sub pages) ... I am lost in never-never land.
So where I used to be #1 & 2 for (let's say): fuzzy green widget ... no problem, but fuzzy green widgets ... I'm outta there and all the directories sites and really good spammers are in.
Anyone else finding the same?
hey, I've tried eevrything else.
My clients site that was #1 for a search for Los Angeles widgets,now has dropped way off of the first page. However a search for Angeles widgets it shows on the first page and a search for los widgets also results in a #1 page result. Also a search for Angeles widgets gadgets also results in a number one listing. Why does the addition of "los angeles" to widgets result in a much lower listing? Certainly there is no use of Angeles without "Los" in front of it and why would Google filter "Los Angeles" anyway.
It is the ultimate money phrase right now and I have a "comedy site" which is filtered out of the results but would otherwise be number 7 for the two word phrase and number 1 or both the three and four word phrases.
I have discovered that searching on the singular form of the searches usually produces the old pre-Florida results or similar to them
Liane, I have just the opposite experience!
One of my sites appeared at number 1 for years straight for "blue widget firm" and now is nowhere to be found.
However...
A search for "blue widget firms" (plural) brings it in at number 1 where previously that search was way down the list.
The interesting thing is that Most of my inbound anchor text points with "firm" AND the title, as well as the H1 tage both have "firm".
The words "firm" and "firms" both appear exactly twice on the page.
This suggests to me that Google has definately targeted either:
1. Title / H1 tags as the same
2. Title / H1 tags with same anchor text
Possibly some combination as well.
I have targeted the plurals throughout my site and in the headers. I think that has more to do with it than anything else.
search for "blue widget firms" (plural) brings it in at number 1 where previously that search was way down the list.The interesting thing is that Most of my inbound anchor text points with "firm" AND the title, as well as the H1 tage both have "firm".
The words "firm" and "firms" both appear exactly twice on the page.
you just opened my eyes to something bobby. I am going to do some tests now.
Eventually, Google will realise this is stupid. As I wrote in a new thread that never appeared, pages do not change between being spam and non-spam millions of times a day depending on user searches.
Absolutely right on!
A major problem for Google's engineers though is this: Unless we narrow down the terms we are filtering we cannot possibly have the processing power needed to filter everything for every permutation of possible searches.
I believe they are using a probability theory filter, if that doesn't sound to much like "The Hitch Hikers Guide to the Galaxy". The problem with probability theory is two fold; 1. It is theory and 2. It uses probability which is far from perfect. I've seen reports that state that Bayesian spam filtering when used on emails can be as much as 99.7% in other words it produces 3 false positives out of every 1000 messages filtered. Apparently this is acceptable for emails and from my own limited experience of using services which use this type of filtering I would agree. However if you work out those figures based on a sample of over 3 billion web pages you have over 90 million false negatives.
If I were designing a system to filter out Google spamming I would try to work out a way to improve the probability beyond 99.7%. Here are two ways to do this the first has athe major advantage of helping the engineers by reducing processing. Actually they both amount to much the same thing.
1. Disregard all of the areas where there are negligable problems with spam.
2. Identify areas of the web in which there is definately a problem.
Some bright statistician at Google has noted a strong corelation between either adwords or the top commercial terms used in searches and the spam problem. So they can do 1. and 2. at the same time and target their attention on a relatively small range of terms to filter. In probability terms this improves the accuracy beyond 99.9% when you look at the total sample of over 3 billion. Senior folks at Google would be well impressed with these figures and would be happy to sanction switching on the system.
What they failed to appreciate was just how much the false positives would be scewed towards the top 100 results. Whilst false negatives across the whole sample might be less than 0.1% in the top 100 for the terms selected I would estimate that it is more than 5%. Since only the top 100 matter at all this is completely unacceptable. When you look at the top 10 results I estimate (certainly in my niche market) false positives amount to over 50%.
Google's selling proposition is that it provides really good search results. What this means is that for most searches the best, most authoritative web pages appear in the top 10 or 20 results. If 50% of the best, most authoritative pages have been filtered out as false positives how can this possibly be of help to Googles future?
Best wishes
Sid
PS The stupidity of the selection of keyword corelated to spam as stated in the passage I've quoted at the top of this post coupled with the analysis I've set out makes this whole episode a complete and utter failure in my opinion.
I suspect they will run with this new index to see if peoples searching habits change. By that I mean people may search at the moment for "widgets" and get poor results, but then refine their search for "blue widgets with pink spots" and get good results. If people do that, then the strategy works.
At the moment Googles biggest problem is the simplistic way 99% of internet users search. I did some research by asking 'inteligent' friends to search for a new job on the internet. Most went to Google and started by doing a worldwide search for "Jobs".... and I thought they were reasonably clever people! They then refined the search when they realised this was not producing good results.
If the way people search becomes more enlightened and sophisticated, producing accurate high quality results becomes a lot easier and website content becomes king again rather than seo targeting of simplistic phrases.
You can actually see this in action. If you do your search for your phrase eg. blue widgets , then do a search for this same phrase but add a minus phrase nonsense text eg. blue widgets -gfhdgytyhf , and you see your ranking are different then you've been hit by this OOP. (Google may change this soon so the minus phrases searches reflect normal searches)
Whether it sticks or not is a different matter.
Chris
Thats the most sensible analysis I have heard so far.
Thanks
If the way people search becomes more enlightened and sophisticated, producing accurate high quality results becomes a lot easier and website content becomes king again rather than seo targeting of simplistic phrases.
Unfortunately, with the current algo, in order to get good results it would seem that you may have to leave out some keywords rather than add more. In other words, users don't need to be more sophisticated, they need to be psychic.
We have a situation where adding keywords to a search (to make it more precise) may actually result in the sites you are looking for being filtered out (as spam?). As I said in another thread, anyone who thinks that is a good idea is certifiably insane.
The definition of a spam site cannot possibly include search terms (i.e. it is set by webmasters not users). It would appear that Google have decided to detect (and filter out) spam sites dynamically according to correlations between search terms and the sites. As a concept this is utterly ludicrous.
As for money terms being picked on - it's possible but I doubt it. The other night I watched an episode of Red Dwarf (UK sit-com) in which a pleasure GELF (genetically engineered life form) featured. Each character saw their perfect partners (rather than a monstrous green blob). This is probably what's happening in reverse. Everyone who suffers thinks they are special and being picked on (their money words have been blacklisted).
Kaled.
that fits in better than all theories of google being crappy .. thats very clever in terms of business if its really true.
MHes said:
If the way people search becomes more enlightened and sophisticated, producing accurate high quality results becomes a lot easier and website content becomes king again rather than seo targeting of simplistic phrases.
Ironically, Google is their own worst enemy. They've pumped up their brilliant ranking for five years now. Think you're having a heart attack? Don't call 911, punch up Google.
The "I'm feeling lucky" button is a tribute to Google's continuing stupidity in this regard.
1. User enters search term.
2. Full result set pulled out of index.
3a. If search term not in filter list apply normal algo and return results to users browser.
3b. If search term in filter list compare each potential result against template of what a spam site looks like. For all sites that pass through the filter apply normal algo but for those that do not pass through assume they are spam sites and DO NOT APPLY ANY ALGO. Then return results to the users browser.
That way only sites that pass through the filter will be given any weighting for those words in the usual SEO places on the page, links in etc.
The template of what a spam site looks like could vary for each search term in the watching list or dictionary or whatever you want to call it. Over time the engineers at Google could refine the spam templates for each search term targeted and teach the system to be more accurate in its filtering.
This would explain why some highly optimised pages do not trip the filter and some relatively un-optimised pages do. Also there is no one answer that we can all share a bit of because the target is continually moving. If this was not the case Google’s strategy could not possibly work long term as spammers would very quickly learn how to play the new algo.
In this hypothesis you can forget trying to find a new algo what you need to find is a template match of a spam site for that keyword term and make sure you get through the filter or make sure that you evolve your site into something that can survive in the new environment. Either way I guess that it could be argued that overall the quality of results and through osmosis sites should improve over time.
If this worked it would be a very good thing IMHO and Google could claim to have led the way.
Unfortunately spam filtering of search engine results is fundamentally flawed. Even if the rate of false negatives was a tiny 1:1000, which is unheard of in any of the research I have seen on the subject, this can a does lead to a very high rate of false positives at the top of the pile. The reason for this is that sites which previously appeared in the first 10, 20 or even 50 results are self selecting in terms of pushing the optimisation boundaries. This means that false positives are going to be incredibly heavily scewed towards the top 100 in SERPs. If your site was competing against spam sites and doing OK then there is a high chance that it might fit the filter template even though it is entirely innocent and ethical. Even with an unbelievably low 1:1000 false positives overall, the figure for the results that really matter might be 50% or maybe even higher.
What the Google engineers need to find is a spam filter which produces less than 5% false positives out of the top 100 results for that term, then they will have done something worthwhile. There is absolutely no point whatsoever in being satisfied with 1:1000 for a search which returns 300,000 results if that means that 50% of the results that matter are missing.
This is the fundamental flaw in the statistics, of web page spam filtering. Only at the very maximum 100 of the results that are delivered matter. Those 100 are the needle, the rest is just haystack.
Google engineers I think it’s back to the drawing board. Switch spam filtering back on when you get less than 5% false positives thrown out of the top 100 until that day you will be failing both users and webmasters.
Best wishes
Sid
Is it even possible to separate the Mom and Pop sites and legit business sites from the junk? Can it be done without massive collateral damage?
Probably not. The most intelligent thing Google did was to restrict this filter to English-language ecommerce sites. This itself was perhaps a recognition that there was going to be collateral damage. But if they knew that some Mom and Pops would get zapped, it seems to me that their timing is pretty stupid. You should wait until after the Christmas shopping season.
Google also needs a structured appeal system for webmasters who have been hit unfairly. They don't like to do anything like this because it's not scalable, but what if they've reached the limits of software already?
One person at Google said that there haven't been any changes in the Google algo for the last seven months, and webmasters were getting complacent. Well yes, and it was getting really easy to spam Google over the last seven months, using anchor text in backlinks. Basically, Google has been fundamentally broken for seven months. PageRank has not been working like it used to. I won't drag up the 4-byte integer argument again, but I still think it's a very strong possibility.
Now we see a rear-guard action to reverse the deterioration of the last seven months. Instead of improving the situation, it has made it worse in the area of ecommerce. The only smart thing Google has done is to limit the damage to ecommerce sites. They didn't even lay a hand on blogs -- take my word for it, because I've been tracking the daily page views on 100 blogs for six months, and nothing changed with the Florida update. They aren't even talking about the update on the blogs, because no one has noticed anything.
The top three blogs about that vanished (83 of 100 missing) technology topic have been removed. You can't find them any more unless you have them bookmarked. Those are run by three different people who don't even know each other.
Technical competency is being punished.
