I think most sophisticated analyses of the blockchain would fundamentally be probability based.

Lets say a crime is under investigation, and one wants to find out who is behind a certain transaction, or alternatively, to find which transaction a known person is behind if one already suspects a certain person. Then combining known "meatspace" data with blockchain data, one can compute probabilities for different scenarios. For example is one suspects that a transaction has taken place in a certain time interval and that the transaction has a certain size, one can reduce the number of suspected transactions on the blockchain, and in the process one gets a more "spiky" probability distribution concentrated on those transactions.

Also, lets say one wants to find out to who bitcoin from a known address has been sent. It is impossible to know if bitcoin is sent to another person or if the person has sent it to himself using a new address.  However if coupled with meatspace data and/or the probability that people actually sent bitcoins to themselves some information could be gathered (gaining information is equivalent to changing probability distribution to a more "spiky" one roughly speaking)

The "perfect" cryptocurrency blockchain would be one where no additional information could be gathered from it. Or stated in conditional probabilities:

Prob(hypothesis | meatspace data and blockchain data) = Prob(hypothesis | meatspace data)

There is a thread where this is elaborated upon:

https://bitcointalk.org/index.php?topic=1011959.msg10978620