As I've discussed before, the key to success in games where you can change your entire lineup every day is to make the proper adjustments for context when projecting player performance. While important in any fantasy baseball format, this becomes the single most important skill in games where you have the opportunity to adjust to the specific context of each game.
There are so many factors to take into account, that the 'model' you build to predict each day's statistics is never going to be perfect. For example, are you going to use weather as a factor? If so, does temperature matter? Wind speed? Wind direction? Precipitation? Humidity? The odds are, you're not going to bother with any of this...although each factor that you include (if you make the right adjustments) is going to make your projections every so slightly more accurate.
One area where I generally try not to take shortcuts though is that I will generally make my adjustments by component, rather than simply taking a player's projected value or points and adjusting that directly.
To take an extreme example of why this is important, let's use the Mets' new ballpark - Citi Field. Some people (including me) believe that it will be extremely difficult to hit home runs at Citi Field, but that the park is likely to inflate triples significantly. For argument's sake, let's assume it decreases home runs by 25%, increases triples by 25%, and overall decreases runs by 10%. Let's say it also decreases overall points scoring in the format we care about by 10%. Think about how this is going to affect Carlos Delgado and Jose Reyes. Is it really appropriate to just adjust everything down by 10% for both players? Not only will Delgado be hurt more by the decrease in home runs, but he's not going to benefit at all from the increase in triples. Meanwhile, Reyes will gain so many triples, that it will almost offset the loss of home runs. Realistically, the park will probably reduce Delgado's value by close to 15%, while reducing Reyes by less than 5%. The only way to properly reflect that in our model is if we make all of our adjustments at the level of specific component statistics.
The same approach is needed when adjusting for anything...opponent, weather, home field advantage, etc.