The man in the article didn’t strike me as particularly brilliant but rather helped underscore the difficulty of the problem at hand: creating a game with sufficiently close to zero but negative expectancy that still incorporates elements which trick the player into thinking the game is predictable. People are drawn to casino games like slots because it unfolds over time, they can see the potential of winning more often than it is possible (slots will have near-wins a far greater percentage of time than randomness would suggest — this drives the addiction and profitability of the game).
Either that, or the games makers are lazy and stupid. Why not use the techniques they have applied on the visible parts of the game and randomly distribute them over a set of already created results-cards rather than having the visible aspect some function of the results?
A good lottery could use stock market data (such as a historical chart, predict the next move!). This would be full-proof as long as the makers were able to collect a large enough set of independent and high resolution samples — maybe even splice samples from different stocks and time periods in a random fashion.