Marc, I have tried to develop a system which simultaneously attempts to achieve four separate objectives:
(1) consistently follow the intermediate trend up on the long side
(2) switch to cash/short before short-term large drops during the intermediate trend up
(3) consistently follow the intermediate trend down on the short side
(4) switch to cash/short before short-term large rises during the intermediate trend down
Right now, for the first time ever, I am optimistic that most or all of these objectives can be (imperfectly) simultaneously achieved. The math of compounding clearly suggests that the bias/preference should be to trading on the long side. Yet, not recognizing the intermediate trend change to down in time will invariably significantly reduce the gains on the long side.
The most difficult aspect is determining the periods of intermediate trend change and trading during those periods. The backtesting of the system was done using only NDX daily/weekly data for 1997-2003 so it was limited in scope/validity as intraday data were not available to me. Undoubtedly, the system will evolve in time as new data come in and yet, surprisingly to me, a very limited number of indicators seems fully adequate to have a very robust system.