Regarding the probability of mateo raping the bankroll I fear that it may not be "so" low.

The website was using half-kelly and standing to this link http://thehackensack.blogspot.fi/2009/11/half-kelly-bet.html the probability of losing, at some point, a percentage of the bankroll follows a quadratic function. In practice my initial invested amount was 0.11335743 and after all the mess I lost 0.09500235, so 83% of my investment. The probability of this to happen was roughly about 17%^2=2.9%, an acceptable number.  (Anyway, it would be great to have a usable model for making these kind of computations, instead of pointing the thumb.)

Regarding the "whale", we can argue that the probability of finding a gambler risking 1 BTC per play is NOT low and also the "timing" argument can hold in reverse: since the bankroll was rapidly diminishing a gambler could have felt compelled to play before seeing it disappear.