I modified an indicator posted by iwjw (a large number of post back). The original indicator tells you when you've got a signal. I added advisory stop loss and take profit lines (they aren't 100% accurate but help with decision making) based on Paulus rules.
At the risk of getting flamed ... I don't think this system should be 100% automated yet (I know that some are trading it that way). The bias between long and short is almost 50:50 and when you look at higher time frame price action the bias is often counter to this. In my opinion this is why the failure rate is on the high side - pushing the long term expectancy down. Some evidence of this can be found in the afore mentioned post (#1454), although I note the sample size is small). Before turning this into an automated system I think additional work needs to be done on bias selection (hence my concern about getting flamed given recent posts). So a PA rather than simple price based filter is probably the next step to improve the bias indication (in my opinion). The challenge here is PA is hard to boil down to a binary value (long/short) - it has a good element of discretion to it - plus the clock used by your broker affects PA interpretation.
Just some thoughts.
ninety47