Question: Has anyone done this before? Is it possible? And would it even be helpful, or completely useless?
I guess that's three questions, actually….
I've utilized a variation of this myself. This centers on non-doubles/non-triples boxes only, in the Daily 3 game (California, which uses an algorithm method).
1) First, you take all 120 possible boxes, running in lowest-to-highest order. (Here's one site that can easily calculate and lay them out for you, using letters: http://www.mathsisfun.com/combinatorics/combinations-permutations-calculator.html )
2) You divide them into 15 sets of 8 ( = 120). However, HOW you divide them, is key—back to this in a minute.
3) You basically track them like you track anything else, seeing what set of boxes is due or not, over a series of plays.
But here's where a math expert's required, not me rummaging around: How would you divide the sets of boxes, PERFECTLY equally? All sets would contain the same number of different digits, or as close as possible; everything would be perfectly even, but not so perfect, or pattern-inducing, as to seem like groupings: all would appear to be a jumble, but would actually be as even a distribution as possible of all similarities in each set. (Maybe there are mathematical models/formulas/theories/etc. already, for doing this?)
The reason for this, is so that the game can't—or as little as possible—create anomalous "long outs," that cause players to chase endlessly & needlessly after "dues."
The overall purpose would be to, as much as possible, limit the variance between each one firing off, having one of its 8 boxes hit. It would probably be best in only pursuing "outs": but the pursuit, done right, would be drastically limited (right?).
I figure 8 is the perfect amount of such sets: not too many as to be unwieldy, not too few as to create endlessly long "outs." KEY, most important: totally "mixed up" so as to not allow the algorithm to escape to groupings and trendings and other evasions.
…? Useful, or just stupid?