And then there was LTCM with the "best" Nobel prize winners. That didn't go so well in the end if I remember correctly. And the world's number employer of PHD's is The Fed and they can't accurately predict what the economy will do in the following quarter.
Looking like Trend will be a simple answer for a few months judging from todays action. DOWN DOWN and yes more DOWN!!
Manual, I could code the Entry and might soon as it's fairly clear cut, but the Exit that's the tricky part, why everyone focus's on Entry and ignores exits. Obviously you could set a 10pt Take Profit, but then you don't get the great risk to reward ratio. Manual is fine for me, don't really believe in auto trading, if possible we'd all be millionaires.
seasonal_decompose in statsmodels for python can infer trend and seasonality quite trivially. If you mean stationary by "non trend" then there are many types of unit root test that can be performed. Augmented Dickey-Fuller as an example.
Advantages and Disadvantages, machine can't think, this looks wrong and get out small profit, but how I often do I do that and it keeps going ?? so tricky! Good Trend moves today and called pretty well aswell, some days every trend entry turns and loss after loss, hard for me to avoid that, hard for a machine, but atleast the machine will keep repeating, not get emotional and screw it up. Tricky!!