I can't figure out how to make a backtest that is acceptable.
For example, in june i've done a backtest of 3 months with my robot ( pattern recognition, trend following...etc )
Timeframe is 4h
i put it on real and in 2 months ( with my supervision ) in august i've payed my holidays....
then....in september, nothing go right. i see that my robot try to open something that i don't like so i stopped it...added some filter ( support and resistance check ) etc...in november i've done a 6 months back test....very good, i put it online....2-3 operation very good....later, all bad. i stopped it...
now i'm preparing for new year....but. all backtest, if i do a test of december...then fails. only if i do a backtest of 4 year in h4 some test is good in december but bad in other months...
i know that it's impossible to do good everytime...and i know that my supervision is important.....but
based on your experiences
a 6 month backtest is needed, put 1 month online and then another backtest and go on like this
or a 5-6 years backtest is needed and then put online for few months?
August and december is banned for robots?
what do you think?