My plan is to define an in sample period for backtesting/optimizing/curve fitting, out of sample period to test the strategies.
Few Questions please:
1. What in sample period would you recommend? I was thinking 2006-2011, but should I be using more recent market conditions to test the edge? I believe I will get about 800 trades during 2006-2011. Or a more recent market condition in sample period is 2014-2016 ?
2. What out of sample period would you recommend? I was thinking 2011 - 2017. Or a more recent market condition out sample period is 2016-2017?
I have read from others recommending to backtest enough to execute 300 trades. For optimization, better don't go way too far back. 2014-2016 should be more than enough as this should check if your strategy can hold out on political news events like Brexit.
Then for forward testing from the optimized backtest, do 2016 to 2017 and see if it profits.
I do backtest of the market. The method of the backtest I do is manual backtest - setting up analysis from the scratch. It is the type of the backtest that I am capable of in term of resources.
Regarding the automated backtest of Expert Advisor consider these:
1. In the Strategy Tester, if the price model of the Open Prices Only gives a good results, then there are higher chances that other type of price models gives better results. There are also higher chances that live testing gives the better results.
2. If the results of backtest is good in smaller timeframe, then it has higher chances for better results in bigger timeframe.
3. If the results of backtest is good in ranging market, then it has higher chances for better results in trending market.
If a trader does not have the type of protection for trading algorithm that is being provided by the black box, the type of protection that a trader can have in the analysis of the market is the adaption of simplicity and the efficiency of round numbers.