Here is a little trend following algo I overlaid on the latest down trending whipsaw disaster (F). You like?
100 view no replies. I've been paper trading it this week on other stocks and it's pulling in about 1% a day. To anyone who thinks you can't beat HFT, this is based on an hourly chart. I could run this program on a 300 mhz computer. Here is a lower complexity variant of it, based on one moving average:
What are we supposed to comment on? I see a chart with some squiggle lines and color dots. How about some explanation of strategy development?
To the above poster: The system is fairly obvious. I like it but since this is ET, there's always room for improvement . 1. Why not stick to 1 security? In fact....STICK TO 1 SECURITY . Your just giving yourself headaches optimizing for 15 diff things unless liquidity is a problem. 2. Whipsaws will obviously be your downfall. It should be optimized a little more if you run into problems with it. 3. I'm not clear on how you define your risk but there's quite a few people that are farther ahead of the curve than you in that respect as well. Overall it looks good though. KISS. But there's more to learn obviously if you feel like taking the "advanced course".
Where are you getting the 1% from? Is it just one week or backtesting average? If it's backtesting, what is the volatility relation to the % profitability? How does it scale accross multiple underlying, if it does at all? Would be interesting to know. If whipsaws are your problem create a trend channel filter. Make it place trend channels around bars and find the strongest (narrowest) ones. Apply signal strength based on channel width and speed (% change per time). Also intra-channel price distribution is useful. It finds amazing support/resistance lines from time to time. Plug it in as secondary input to composite signal and tune to your liking. Trend channels are easiest to debug for generating signals too, and they allow the risk/reward aka money management AI to function more easily.
There are other dimensions to this idea that I didn't mention, but the 1%/day was based on about 5 stocks over this past week. Don't laugh - it all starts with a concept. So far the expectancy is working out to be (6*.03)-(1*.005) = .175, with a win rate of 86% and an R value of 6. This falls in line with a similar system I was using on daily bars, until the Bernanke Bull market broke.
I won't laugh as long as you know that it's just a concept. Using a week or even a month can be very unreliable. Even a 2% daily average on such a short term can easily turn into -4% within the next period. The less trades the system does per day, the more unreliable it is over longer periods. It's best you do a lot of backtesting, and if your system is learning something along the way, then do forward testing also. When I developed my first automated trading system it worked with PSAR and general trend detection (to avoid whipsaw). It was profitable for a few weeks of backtesting I did, and then live it went very negative during the next week. I had to scrap a lot of ideas and start from scratch, so my advice to you is, don't be afraid to throw away ideas. You can also keep some for later testing, so modular design of a strategy is important (ie you can turn them on/off based on market conditions). If you can (if you are a programmer and have developed your own platform), make a time-step chart where it will replay a chart and then output to graphs/chart areas all the signals that go into the buy/sell signal composite. That way you can pick the worst days, see what went wrong, and tweak or introduce new rules. Other than that, you can get a supercomputer and try all the 100 million combinations within days for multiple years that's what I wish I had right now.
Note: if your system is based solely on indexes, trade the leveraged ETFs instead, or VIX futures if it's very short term (intra day) - it will turn a 4:1 margin into 18 or more. For example, VIX on an intra day basis moves tick for tick with the S&P (see attached). So trade VIX instead of SPY