So while attempting to program a strategy which enters upon divergences, I screwed up and entered a buy where it should be shorting, but also a buy when it supposed to buy. The reasoning behind both entries made no sense together, but in the testing phase (before i realized the error), the equity curve actually looked pretty good. So, my question is, would you trade a strategy that made no logical sense as to why/when you enter/exit the market? Lets assume that the system was to your standards of profitability and risk, and everything was tested properly, ie no curve fitting, good sample size, etc, etc.
I guess a draw down phase would be mentally taxing, even if it is tested right. I wonder how long I could keep the faith...
i've had similar experiences. actually, have learned from silly mistakes like that. but to answer the question, i wouldn't trust it unless i could explain the rationale behind it. maybe a long standing bull market for example? bb
Its actually an intraday ES system. The sample size was in the thousands, so I wasn't too afraid of being stuck in just one market environment. Equity curve was constantly good, but it actually did even better in the low volatility period after 2003.
My guess is that you have another "oops" in your code that makes your strategy look good in back test. Post it, we'll backtest for you.
The example I like to keep in my head is: As the number of high-seas pirates have decreased over the years, the average temperature on earth has increased (care of http://www.venganza.org/about/open-letter/). Just because a relationship fits all known data without being hand-optimized, that doesn't mean it has any predictive value. In my opinion, if you don't understand why a system works, it's way too risky to trade it. That doesn't mean you shouldn't spend a lot of time trying to make heads or tails of your accidental system. But, if it defies explanation at the end of the day, I'd suggest you let it go. Either way, good luck!