I trade crypto and try following strategy development guide.
I have generated so far around 28k strategies to trade BTC and only one has passed all recommended tests.
I have data from 2017.01.01 till 2019.06.01.
I use first 60% (18 months) part for strategy generating, 20% (6 months) for oss tests and last 20% (6 months) for final oss test.
Valid strategies output percentage is very low.
What problems I may face if I will use all data available for strategy generating? (60% is, 40% oss) and after that 20% for oss tests and last 20% for final oss test.
There is a problem with old data
In crypto price of BTC for example is dynamic over the years: 0.01$, 1$, 30$, 1300$, 20000$, and pips settings are dynamic too.
I have tried data from 2014 with pip settings from 2017-2019, as the result I got 0 strategies after 24H of generation.
This is probably because of huge slippage because of pips in the old data.
I would suggest new pip setting – significant digits precision (described in the thread above)