I have been developing a number of automated strategies, and thrown most of them out. This is the simplest and also has (it seems) the most robust follow though (ie optimize on two months Jan/Feb, test on 6 months Mar-Sept) but the win rate is 'flip the coin' 50% , fortunately the there are more winning days than losing days (about 2:1) and generally the winning days are better than the losing days, so the sharpe ratio is 3.5 on the testing period is decent, and stays above 3 even with some minor changes in parameter (ie range bar settings etc) on the strategy.
Setup:
3 time frames (5 range is the actual trade entry, and Better Renko(thanks Aslan!) used on the others
5 targets (3 are covered if not hit within 'x' bars, 2 are set to trailing stop/runner target)
5 range bars for the exit & trailing indicators
EURUSD 305 - 1510 eastern time, with (as yet) no 'off time' for news and no 'daily PnL stop' either
ATM, the strategy is only LONG ... and was optimized in a generally bearish time period
There are three things I would appreciate comment on:
1. In personal experience is 5 range granular enough for backtesting in NT7 ? Up till now I generally stayed in the 3 or 4 range bar area.
2. 'expectancy' was used as an optimization parameter: score was 24. Am I right to (now) concentrate more on Sharpe ratio than on the winning% or winning ratio?
3. Personally, would you only test on 'quantity = 1' (as I have done for each of the 5 targets) or would you add quantity as you move forward in testing (ie PnL daily levels) ??