Please correct me if im wrong!!
I think you are correct as you are (presumably) building your System using all default values for the variables?
But what about any parameter levels (correct term?) … how do you decide what value to use when building System on 100% of data?
Okay so lets say i want to make a simple strategy, using moving averages and bollinger cus why not..
I would use standard 2 std and 20period bollinger + the moving average i think would be best. From just looking at the chart i would decide on the values that i would like to use for my theory/strategy. Let say moving average 20 and moving average 40.
I would use those to create my strategy, and if i like what i see i start using 30-50% of the data (Maybe data from the middle part, or the end, or the start..) and i optimize on that data only. Then i just check the 100% data again and if its curvefitted (happened MANY times ofc..) the strategy would just go to shits when looking at anything outside the curve-fitted data… Ill include a picture of a strategy i curvefitted, in case someone would like to see what im talking bout.
Edit: i always leave the last years as out of sample.. whole 2016+2017+2018 is usually always left out.
Edit2: After i check my strategy and i start optimizing on 30-50% i NEVER EVER change the strategy. I will optimize the variable, but i will never add indicators/other stuff to the code (only exceptions are “what time to trade” or “sell on fridays” or trailing stop loss) but i will never add like “average true range > XX” or something like that. If its not in the code when i start optimizing, its a completely new code.
Edit3: If the strategy looks like the one in the picture, i throw it away 10/10 times. I will never start to re-optimize or add stuff to make it better… When i optimize its do or die-. Either its good and i run it in demo then live, or its shit and i discard it.
