Been thinking about setting bracketed stops [up and down] on a stock position based on probabilities of the stop price occurring and the profit or loss involved. Figure this is a volatility-based model with a trend and/or 'TA indicator' bias developed from regression analysis and folded in to the probabilities.

The key would be to keep score and adjust the stop level guidelines accordingly. The obvious guideline is that your stop level brackets should not be set such that the odds are against your making a profit over a number of trades.

Building the data to back-test this model, and then to simulate in real time, will take a while; but I figure it's a worthwhile pursuit. Maybe Valentine's Day for some prelim results?