Found it... It's not volatility. I can remove the effect and returns look better Please, don't get confused, we all know it isn't going to work in real life.
if you put a moving average on the number of tweets, they will go from essentially zero through Nov. 2016, to something rather large. If you put a moving average on the returns, they head downward *perhaps* through Nov 2016, but rather plainly over the last 2+ years. (to my jaded eyes.) (((Huh. I'd check out a correlation to TWTR prices as well!)))
I've figured out that the main reason this works is because of alternative data. WTF, never thought that shit would actually work. Don't worry the "Found the bug" post is coming eventually.
You never know when currency pairs that performed best in a backtest over a given period will start to perform the worst in a future period. The currency pairs that performed best during that period did so due to a multitude of domestic and international economic and political factors that affected BOTH currencies circa to that period. In most cases, backtests are unreliable in determining a best currency pair or instrument for trading; you're best-off focusing on trading the ones that are the best now.
It's precisely because I don't know which will do well that I want to trade them all. This week I'm busy but next week I should be back at it
The trader's currency trading strategy is usually made up of trading signals that ... For example, they may notice that a specific currency pair tends to ... A forex trading strategy worksreally well when traders follow the rules.
I've rewritten this test three times using different approaches (vectorized, event based, "tick by tick") and finally found a bug today. Except the bug was that I had the conditions incorrect. I think I'll probably start paper trading this one.