When I first tried to do a trend follower model, I took the simplest approach with the rules:
1) If Bull/Bear indicator is Bullish, and no scalp is active, then go long.
2) If Bull/Bear indicator is Bearish, and no scalp is active, then go short.
But this was getting me in trouble. After intermediate bottoming events (like the most recent), the Bull/Bear indicator could stay in Bear mode for a week or two before cycling over to Bull. During that week or 2, the rules would have me stay short while a rally kicked off. And I could miss (actually be short) the juiciest part of the move.
If I had a decent model to identify intermediate bottoms, then I could override my own Bear mode and go long. But I think it is difficult even for the experts to identify Intermediate Bottoms (and tops). I tried and gave up.
Short term bottom finding is a little easier because you have much more historical data to work with. I average 25 scalps per year. From 1996, thats alot of scalps.
But with intermediate bottoms, you only have 1 or 2 per year. You could generate a set of rules that works against your history of 15-20 IT bottoms, but curve gitting comes into play and you may enter the next bottom too early or miss it entirely.
Currently I am using someone elses model for trending positions, but I am not really happy with it.
Market Data 
Markets 
AI
