Thoughts on Strategy: Voting strategy is susceptible to the Tinkerbell effect. Certain conditions only exist because voters or their programmers "believe" them to.
Coming up with voter strategy piecemeal is not methodical enough. I propose a fairly reasonable alternative to this. I call it Vote By Result. I have a general purpose strategy idea. Divide the voters into particular groups that vote early to late. The earliest group would presumably vote honest (I recommend making it small to minimize the impact of "false honesty") The others would abandon their true favorite if they lacked the power to independently make it reality. Essentially after each group goes the results-so-far are updated. Then each pseudovoter* in the new group votes. The process continues * a pseudovoter is like a block of voters, but its vote is given more weight than an individual voter and agrees with itself completely so that voters in later generations can counteract the "momentum" of previous results. The result of this experiment would be fairly simple. The voters would end up using simple strategy. Later pseudovoters would abandon their true favorite if they do not possess the power to independently alter results and instead to default to the best electable alternative. I have no idea how the weight/number of pseudovoters should grow with "time". I suggest geometrically growing pseudovoter weights and a constant population size. This method is computationally intensive, so beware. Probably not as intensive as generating theorems proving some strategy optimal for some voting method however : - ).
---- Election-Methods mailing list - see http://electorama.com/em for list info
