Whenever I try to find strategies on smaller timescale 5m,15m, the strategies development per hour drops dramatically.
One solution was to shorten the timeframe from 10 years to 1 year.
However, the found strategies failed in the backtest.
My idea is the following, a tool that mixes the data of 20 years.
MIX the historical price development by random generator.
For example,
Month 1 of year 3 month 1
Month 2 from year 8 month 2
Month 3 from year 1 month 3
The goal should be to reduce the amount of data to 20 percent.
Would that work?