I've always wondered about this but never got around to doing any experiments on it, so I'm curious if anyone else has any insights. Consider a hypothetical situation as follows:
--A team will play half of its upcoming games in a park with high home run park factors, in the 170-190 range. The other half will be played in a park with very low home run park factors, in the 40-50 range.
--The team has two part-time third basemen who each have enough available playing time to play about 50% of the season. So if it makes a difference to the team, one of them can play all the games in one of the parks and the other can play all the games in the other park.
--The two players are equivalent in every way except one: One of them is a power hitter (power rating of Ex) and the other one has average power (rating of Av).
My question is: Does the team maximize its run production by playing the guy with Ex power in the bandbox high-homer park, or by playing him in the cavernous low-homer park?
I can think of logical reasons why either approach would be better, depending on how DMB models power. Power hitters are generally more likely to hit more balls for a long distance, AND when they do get a long hit, to hit them them further than an average hitter. If frequency plays a bigger role than distance in DMB's modeling of power hitters, then I suppose it is better to put the Ex power hitter in the smaller park, so that you can benefit from all the extra long hits he would launch toward the short porch.
But if the distance of batted balls plays a bigger role than the frequency of long hits, then I could see why it might be better to put the Ex hitter in the bigger park and the Av hitter in the little park. The reasoning would be that if a 310-ft. fly ball down the line is going to be a homer anyway, it doesn't matter if the Av power hitter hits it 325 ft. or the Ex hitter belts it 420 ft. But if it takes a 400-ft. blast to get the ball over the fence, the only guy who is going to do it is the Ex power hitter so he should play in the low-homer park.
Has anyone done any experiments to test if either theory is more likely to be correct?

