First, I would like to explain the history of SEOs and consequently why many SEO books do not provide the exact implementation details for a general audience.
From my understanding, there are very few people who know exactly how search engines work. If someone knows how Google's search works in great detail, he/she can make a lot of money by discerning how to cheat the search, and (for a profit) ranking someone's website into the top search results. This is why Google/Bing's searches are proprietary rather than open access.
In addition, the search engine is necessarily a very complex system. You can expect any decent search tool to make use of nested algorithms, thousands of hand written rules, AI-assisted human filtering on content, and so on.
In summary, while components of the system may make use of well known learning procedures, the overall system cannot be described simply by math. Even if that were possible, the system is sure to reflect newer and more sophisticated features by the time any one person came to grips with its current state. We do not know how often Google update their algorithm, and what exactly those updates entail.
Since no one person knows exactly how things work, some observers conduct experiments to see or modify results. Some tricks are really hacks. For example, some site designers include "dirty / hot / unrelated" key words in the page, but make the text transparent so browsers clicking-through do not see the text although that text essentially swayed the search that originally sought such content. These types of tricks are falling out of favor, and are examples of how Google's dynamic algorithm development quickly matches and overcomes the site designers' biasing efforts.
These are examples of SEO tricks, and illustrates why search implementation quickly deviates from a purely mathematical or statistical treatment.
Above are real world examples of SEO, and why there are not too many math in the SEO books. But if you want to study recommendation system in general, there are many places to start with. Such as
Learn to rank
Non-negative matrix factorization
Eigendecomposition (Markov process, stationary distribution / original page rank algorithm)
And many more.