It doesn't make sense to me either.  I think I either made a mistake, or there are errors in the file I downloaded from dooglus. 

If you look at Nakowa's actual bet data, he makes a bunch of small bets mixed in with his large bets.  If you take his actual data, but just shift one colum so that he wages what he actually wagered 2 rolls earlier, then most the losses line up with the small bets and most of the profits line up with the large bets.  This should be statistically impossible.  So either I made a mistake, there is an error in the file, or the seed was known by Nakowa and he left a clue for us to find.