I've been developing a few new models recently. Last night I was working on a relatively simple trend following type strategy and testing it on the 1-hour EURUSD intraday data.
What I discovered was that from 2007-current, the model run as "long only", the equity curve printed elegantly, code ran robustly and very profitably (at 2% risk per trade it returns 380%), trading 24 hours a day.
When I included short sales in the same (but inverted) setup, the result was pretty dismal. Still profitable, but short sales were a net loss (at 2% risk per trade it only returns 150%).
I tested a few concepts out at this point, and discovered the system could be returned to profitability only by modifying the exit condition for short sales so that they were much longer term (total long and short at 2% risk per trade return 580%) - or simply not shorting at all.
Basically, the original setup calls for a relatively tight trailing stop that seems to be highly effective for longs. However the setup on the short side only seems to work with a much much much looser trailing stop. This then completely changes the stats of the system from a relatively high winrate/low DD system to one which can take a 46% modelled drawdown and 16 consecutive losses in a row. Using the loose trailing stops for longs is not effective.
Now, I am not losing sleep over this models behaviour, I have written code which can already successfully capture both long and short intraday trends in the EURUSD 1-hour.
But this result surprised me, especially considering there have been some legendary intraday downmoves in the EURUSD since 2007! So I thought I would come here and poke the brains of ASF systems traders to hear their opinions.
Can the system be reconfigured so that the exit condition for both long and short is the same and remain profitable? Or is this just a long only system? If so, can you explain how that works on an instrument like EURUSD which is theoretically a pairs trade and therefore theoretically just as shortable as longable?