I am at the point in my research and testing where my minute level pairs trading algo more or less does what I wanted it to do.
enters positions when there is a reasonable expectation that spread series is stationary, and exits when balance in spread is re established.
at the moment though i dont have any stop loss logic in it.. when spread does not come back, algo keeps positions open until the end of the day, sometimes it recovers, but sometimes it incurs big losses
the question i had is about the logic of stop loss that other quants are using for this kind of algo.
i was thinking about putting a simple one: if unrealised loss passes a certain threshold, close all positions
or this can also be tied to a certain threshold zscore value to stop loss.
I also noticed that majority of successful trades are quite fast w.r.t. time difference between entry and exit. whereas unsuccessful are much longer. did you try to exploit this property for stop loss logic?
another one could be use of stationarity test, when it is failed, close the positions..
keen to hear your experiences and ideas on this topic.