This would be done by allowing strategies to be tested on multiple assets where trade hedges or correlation pairs trades would occur simultaneously between assets. In hedge mode, you can pair two assets against each other like EURUSD & GBPUSD. The genetic algo search would look for opportunities based on the custom fitness of strategies where either of the following was true: a strategy is identified on either asset where buying GBPUSD and selling EURUSD produced a profitable trade record or where selling GBPUSD and buying EURUSD produced a profitable trade record. In correlation mode, the same logic would apply but it would be looking to trade both assets in the same direction. In basket mode, it would be the same thing as hedge mode but with more than one asset and the directions can all be independently changed of the other assets. If you dont want a mode for these, you can just categorize everything from a directional bias instead (load all assets to look for buy strategies on in one section to compare them with sell strategies in another section. If one section is blank and another has multiple assets, its correlation by default) If there are more than two assets, the genetic optimization should be able to figure out if adding any additional asset is necessary. So if looking for a strategy between EURUSD, GBPUSD, USDJPY and GBPJPY, we should be able to see strategies that just work with two out of those four assets, strategies that only work with three out of those four assets and strategies that only work between all 4 assets.
Stop loss/take profit can be based on the exit rules of any individual asset, the ATR or pip based tp/sl of any asset, or the difference in spread between all the assets being traded. (please allow fuzzy logic to account for multiple situations possibly occurring for entry and exit) However, all assets would be opened and closed at the same time.