Jameson Quinn wrote:
3) Create (by hand or using genetic programming) and test "strategy heuristics", by which a given voter can use the polling (and knowledge of the underlying probabilistic models) to estimate the expected value of various strategic options, assuming all similar voters use the given strategy, and assuming no counterstrategy or defensive strategy.
Could your third point be done, for very small electorates, by use of minimax game tree algorithms like the one used in computer chess? The objective for each voter would be to get his own candidate to win (and for rated methods, to have the winner maximize his individual utility). Minimax requires perfect information, so that's a flaw, but it should give a bound, as it were, because the voters can't know more than perfect information, only less.
A variant to model differing capacities of strategy may have different voters consider the opponents' moves with differing horizons - very good strategists consider "A's move to B's move to my move", whereas not so good strategists consider only "B's move to my move". It might turn out that in a minimax case with equal horizon for everyone, the method becomes some kind of DSV and thus doesn't do bad at all - whereas in the real world, not all strategists are equal.
---- Election-Methods mailing list - see http://electorama.com/em for list info
