Which method are you using to determine statistical outliers.... ? Market Profile... VWAP....Trendlines?... ATR's?....
The outliers are 3/4 sigma events on the normal distribution, these days everything is perfected close to +/-0, the issue is that are always leaks. You want to plug the leaks around the perfection to make sure that nothing seeps through causing a chain reaction. Well if the institutions can't do it with GBPUSD this week, LTCM couldn't do it, you can be quite sure no one else has a clue. So you trade the margins and hope you're not in the market when the leak ruptures, or you trade the outliers and watch the marginal gains from the sidelines. Then there are the handfuls who can do both, but you really don't want to go down that route.
http://scikit-learn.org/stable/modules/outlier_detection.html http://machinelearningmastery.com/how-to-identify-outliers-in-your-data/
Machine learning won’t help as it will quickly curve fit stops to past optimums which you can actually prove aren’t relevant(you just got lucky for periods at a time).
when I meant outliers, the actual price chosen is of significant importance for entries. So how does one base their entries in a specific derivative. So basically your looking for temporal outliers in price.