I guess I shouldn’t call it a “major overhaul”, but that’s the simplest I can put it. Originally I intended on simply adding or re-tooling some of the price conditions my program looks at, and adjust some of the thresholds. But after going through a lot of my code, I’ve found other things that can be improved, and some bugs that had been hiding for a long time. So for the time being, I’ve stopped trading altogether until my program once again reaches a point where I’m confident to use it again.
In a sense, this is becoming a major overhaul. Seeing the program in action for a few months, seeing the results and noticing things in general (both with my program and the market), I’ve found a few things that could be vastly improved. For example, the code I have to quantify what a price is doing; since I’m using strictly numbers and no imagery (no image processing, chart reading, etc.), I have to come up with clever ways to “simulate” charts. In order to try to determine in which direction a price is moving over a given period of time or to “see” chart patterns (“head & shoulders”, “tea cup”, etc.), I have to figure out how to mathematically describe those movements. Higher math was never my strong suit, so I’ve had to rely a lot on actually drawing things out and coming up with numeric representations for what the lines are doing. Like coming up with a way to represent an angle as a percentage, for example.
A lot of how I’m doing things may be arbitrary, but ultimately what matters is that the results make sense, and that applying the system to random stocks yields expected results. So far, so good.
I’m also working on adjusting my scoring systems, so that they’re easier to understand and interpret. One of the things I had noticed as I was starting on this overhaul was that a lot of the scores simply didn’t make sense, in particular some stocks having negative scores when results were favorable. The key now is to have “targets” integrated into the scoring, what I would like to get out of my trades. So if I want to average 5% per trade, a stock averaging 3% will score lower than one averaging 7%, naturally. I’m a little concerned that I’m either over-simplifying or missing something in my new scoring method, but so far it makes more sense than the previous one.