Machine Learning in trading - page 2

[Nor Azman Bin Ramli]

Rather than focusing on continuous optimization, it's a good idea to measure your robot performance (if it's still work or vice versa).

If you optimize your parameters for 2019 data period, and then you backtest your strategy for 2020 data period, it still give you a different performance or maybe worst.

The important point here is if you found profitable patterns that has statistical edge, measure its relevance over time rather than optimization.

Again, this is just my opinion.