Is it possible to have different costs definition depending on entry or exit

[tfrojd]

I am working on a strategy that needs to take position in auction market and scale out of the position in the continous market. Is it possible to have different cost definition for entry and exit in some clever way?

[enzbus]

Hello, yes it can be done. There are 2 types of costs, SimulatorCost used by the market simulator, which implement this interface, and cost objects used in optimization. For those it's a bit trickier since the method called on them at each iteration does not return the cost itself; it updates the value of CVXPY parameters of their expression; it is this method. Both methods receive the current weights of the allocation so they can implement logic to make the cost dependent on your position, whether you're getting in or out. If you derive from SimulatorCost itself it can be possible to define the same object that works both in simulation and optimization, but you should be careful to create different instances to pass to the policy and to the simulator. I'll try to improve the documentation of those, ideally the manual page should explain how custom cost (and costraints, ...) are created.