Throw an ice cube into the beginning of a class iv+ stretch of river. Now, write an algorithm that predicts (in real time) the magnitude and direction of every movement of said ice cube. Now, apply that algorithm to a bucket of randomly sized ice chunks and apply a transformation to each movement that represents taxes, fees, etc. and accounts for real-world issues like network hops and liquidity.
Or, you know, give up on that and write an equation that outsmarts the other guy's equation.
Of course, I haven't spent enough time on this to know what I'm talking about (I've found that once I reach a level of competence on a topic I wouldn't dare blog about it because I realize how little I actually know about that topic) but it is fun to think about and yep, this is what I've been thinking about lately.