A: Right! This was hasty of me to phrase it this way. I was thinking about it today a wee further, but the coverage is too imprecise to make me understand how the sudokus were produced (and how the unconstrained matrices were simulated). If the sudokus were drawn uniformly over the set of all sudokus, the entropy should be where *N* is the size of the set of all sudokus… Since the set of *all* matrices is much larger, , we should have . So some information is clearly missing!

I’d say it somewhat differently: It is the _procedure_ for creating the sequences, not the sequence itself, that is random. When you talk about sequences, certainly some are more random than others. 111111111 is much less random than 1000111000. If you wanted to, you could go Bayesian on this and say that your posterior probability that the underlying process is close to random is much lower, conditionally on observing 111111111 than after observing 1000111000. ]]>