automated systems update: goodbye to the pioneer systems
Goodbye, pioneer systems
It had to happen sooner or later.
I started building automated systems in 2005, and from then till early 2008, all I worked on was these two systems, which I later called the "pioneer systems". They were based on the EUR and GBP, they entered (in the last version of them) from 8.15 CST till 8.45 CST. Their rationale, initially, was the difference in bid-ask volumes. If there were many more contracts being demanded at the last level of bid, compared to the contracts offered at the last level of "ask", this to me meant an impending rise. And viceversa for the opposite situation. (By the way, I still believe this concept has some validity but it's too hard to test it right now: I spoke about it in my previous journal).
These systems were never back-tested because I didn't know how to back-test such a thing, based on the bid-ask quantities. I don't have the data nor the tools. However I was positive that it worked. So I forward-tested it for years. After over two years of these forward tests (and many changes), since the method didn't seem to work at all, except for breaking even, I changed it and turned it into simply this: during the first half an hour of US trading (from 8.15 CST to 8.45 CST, because the system starts 15 minutes earlier) the low and high of the day (futures' prices) work as support and resistance levels, and you can trade off them. This system was easier to back-test and so I did, and modified it accordingly. However, despite the pretty good back-tests, it still is the worse of the 42 systems and it doesn't even break even for the past 7 months. So I had to get rid of it, also because it was the heaviest two mother ****ing systems I had on my excel workbook, which, without them, went from 5 megabytes to just a bit over 1 megabyte. Those two "pioneer" systems relied on price being polled once every minute, they had a chart apiece and each chart was based on a heavy database of those one-minute prices. Didn't work, were heavy... had to go, despite their historical value.
40 "side" systems
The very amazing thing to me is that while I wasted 3 years on these two systems that I abandoned today, in early 2008, I decided in just a few months to build a few "side" systems, so I would add some activity, because these "pioneer" systems were trading only once a week, and things were getting very boring for me. But this time I did things right, and didn't start from forward-testing but from back-testing. I then automated what worked in back-testing. This is how I created what i grouped in 8 waves of systems, between early 2008 and late 2009, consisting of a total of 40 systems. So i am finding it very funny that the two early systems have now been abandoned completely, and I am left with what began as a side activity because I was bored. Whereas at the start I thought my life was going to be based on these two wonderful systems. I guess it could be the same in building houses. Take Boston or some other city. You are pilgrim, you get there, build a nice house, or a
church, which you consider the ultimate building, and then, hundreds of years later, it is
surrounded by skyscrapers, and only being preserved for historical reasons.
So this is what my two pioneer systems looked like before I abandoned them, also in terms of profit (the skyscraper represents the 40 new systems):
Yet I abandoned them, because the only objective here is profit and not preserving historical systems. I felt bad, but it had to be done. I had to get rid of what didn't work, once I was positive that it didn't work.
I got rid of the DX filter as well
The week for the other systems was good, and I've added some changes to those as well. I got rid of the filter which, based on where the dollar index is going, reverses my systems' signals. There's no point in reversing the systems if they are making money. The only filter is therefore the "past week" filter. If we made money last week, we keep the systems running the same. If we lost, we reverse them. I am counting on the fact that the red weeks never come alone, but in groups, and also that they will mirror how the DX moves, so it will be probably be the cause for the red weeks, but I won't have to worry about the causes as long as the red weeks come in groups. Of course, as I said in previous posts, this requires that I keep track of how the systems would have performed had they not been reversed, so when I speak of red weeks I am speaking of red non-reversed weeks. I had to work quite a bit on formulas and macros to take care of this problem.
Anyway, here's the new equity line:
And here's the new profit, week by week, group by group:
And now I can finally go to get drunk or something, because a lot of work is behind me.
