[TheAlchemist]

The real problem is not in finding an algorithmic strategy that seem to work on historical data - the problem is how to validate sure it will work in the future and how to be confident it's still working even if it's loosing money for some time ?

[piyiotisk]

Yes exactly that. Even if you backtest for many years it’s still possible that you’ll lose money because future although correlated to the past can always have different patterns.

Still it’s a really fascinating field to explore.