In 2008, I spent a few good months developing strategies in TradeStation full-time for the index futures. Now after a full year of break in development, I am starting to devote more and more time in quantitative finance research again. In hindsight, the biggest shortcoming with my previous strategies is that they are not discriminating enough. Using an analogy from robotics (my specialty), some of the most basic robots are great because they can do one task, and one task only, especially well. Those developers that are too ambitious in trying to conjure a jack-of-all-trade robot typically fail miserably. Our current technology simply isn't sufficient for anything more than a specialized automation. Much is the same in developing automated trading. With my limited resources and skills, my goal is to build simple automated systems that can do one job only and do it damn well. Once that can be achieved, it's only a matter of concatenating all these individual systems to exploit more and more opportunities. Therefore, the new plan which I'll follow for development are:
- Identify a recurring profitable scenario in the market
- Isolate the conditions which would yield a high probability of such scenario happening
- Program the findings into my platform
- Run backtests to calculate the expected profit (probability * reward/risk)
- Tweak and debug
Ultimately, remember this: It's not a matter of the number of trades you take, it's a matter of being as picky as you can in only playing the best of the best trades you can find.