If Google collected lists of bad links and waited to take action then it wouldn't effect links instantly. This instant action is what Penguin Live is supposed to do rumors have it.
Let me put some pieces together for you that you may be missing. Currently Google does not announce updates to their algorithms, and sometimes does not confirm them when other spot them. The last generally accepted algo update was Nov. 19, 2015. I started the link building mentioned after that on Nov. 23, 2015. This mitigates your assertion that bad links are dealt with after an update.
Your claim of Penguin live would be a reference to the RankBrain algorithm.
https://www.besthostnews.com/ai-rankbrain-now-part-of-google-search-algorithm/
Reading that link, RankBrain is nothing more than a semi-supervised Bayesian Classifier that connects words used in a search query and contracts them into general itemized categories. The problem is that such a classifier can be wildly inaccurate in use and can go off on some wild tangents. Most often, this type of classifier is used with a small input set that is classified and the expanded results are inserted into a new run of the classifier as the training data. This process is repeated numerous times.
Knowing the above, and being able to build a classifier in Python in about two hours, an enterprising person could scrape a large number of webpages, classify the title, headings, and body and make a very good estimate of what RankBrain finds important. Once a person finds the words that the algorithm finds most statistically relevant. that person could write an article that intelligently incorporates the most statistically relevant words into an article. You would scam RankBrain in very short order.
Anytime I start building links, I use a process that is somewhat similar to what RankBrain would use. The difference is that for the last several years I have been flying by the seat of my pants as to what is important. Lately, I have been working on Bayesian classification; what that will work out too, I do not really know at the time. Considering that it takes my I7, 4770k with 20 gigs of ram 17 seconds to calculate every possible combination in a 69 choose 5 matrix (11,238,513 combinations in a 10 x 7 matrix), along with the deviation and variance, as well as three gigs of ram to do do it, you can determine just how computationally and resource expensive it is. I can speed the process a little using my GPU, but not as much as you would expect. By way of contrast, it only takes three seconds to calculate every possible 5 card hand in a deck of cards (52 choose 5) and does not break a gig of ram. Scaling up, you need some serious computational power devoted to search queries. Human language is a large data set.
You may also wish to look at rank ranger while remembering that I asserted that I noticed changes on Dec. 09, 2015
https://www.rankranger.com/rank-risk-index
MozCast also shows a small spike for Dec. 10, 2015
http://mozcast.com/
Algoroo also shows a small spike on Dec. 10:
https://algoroo.com/
Last edited:
