I've been trying to gather statistics on various patterns using NT, but am running into what is probably a common problem - the imperfection of the pattern.
For example, just trying to program a simple 2 or 3 bar pullback raises many problems when it comes to programming. What about bar size, bar shape, nearby S/R lines, strength of the trend, and on and on. After programming a simple pattern and then looking back at the trades, there's a large fraction that I would never consider taking when I look at the context of the pattern. Then of course, there are the others that work that I wouldn't consider taking because of the context.
It frustrates me to not be able to use the power of the computer to gather statistics. It's even more frustrating that after doing this full time for over 2 years I still don't have a trading plan that I have confidence in.