Quote from achilles28:
Try lower time frames. I don't code art. I quantify simple relationships that compose the larger picture. It also helped to include a fail/reverse clause into programs. Another tip: add to winners, Livermore-style. Pyramiding works so the initial trade is only a fraction of the total position. That way, when you're wrong, you're wrong for a penny. Right, right for a pound etc..
Good insights (all of the prior posts you made here), IMO.

The only issue I have as a retail trader, is that while I agree there are many good edges present in much lower time scales (superior to higher, such as autocorrelation, for example), the practical issues, like getting filled, size, slippage, comm. etc can supersede theoretical edges based on back-testing historical data.

