(Note: This challenge will require use of simulators, but also a lot of non-simulator related thinking. I also don't know if this challenge has a solution, but I suspect it likely does)
The challenge is simple in concept: Construct a Dominion board with three 'optimal' strategies, such that strategy A will beat strategy B, strat B beats strat C and strat C beats strat A, and there isn't a strategy which beats all three.
Essentially the challenge is two parts: Firstly making the board so these strategies exist, and second making some simulator scripts which show the A>B>C>A strategies. Note that if two of the strategies are very close (e.g. C beats A 45.5-45) then it's not unlikely that A could just be optimised to beat C, and C might not be able to recover that loss, which would make A simply the dominant strategy.
I've done a little thinking about this, and have the following thoughts. None of them will probably be that mind-shattering, though.
To get the RPS effect, you're certainly going to need some kind of player interaction. Although I'm not sure what kind would work.
A good start might be using attacks and counter-type defences - e.g. a player buying Tunnels might not do well against the Council Room player, but would have a good shot against the militia opponent. Or using cards which turn specific opponent attacks into an advantage.
The main difficulty is probably going to be there NOT being an optimal strategy that beats all three. I reckon finding three strategies in a (rather strong) RPS effect might be a good start, though.