Trading system development can be a great learning experience if it is done right. Great discretionary traders are born. It is a god given gift. Great system traders can be created. Great system developers only need to be one with the market while developing a trading system. The key is every system must start with a premise, a concept. If that is sound the system has a chance of working. Data mining for patterns without understanding why they should work is one reason that many systems fail.
Even if I use a mechanical system I still have to select the securities to trade. I have to decide how much risk I can tolerate. There is always a discretionary component in my trading.
nicholof You are correct.... The progression from throwing darts to being a systems trader is a long road for sure. I think "MOST" good traders get to the systems stage once the trading game becomes rather repetitive with the same signals, rules etc. I myself would code everything and set up "AUTOMATIC" in and puts if this software industry and platfroms ever get to the point of reliable execution. Sounds like we are close, but i will not yet feel confident to sit there and light a Cuban cigar and feel confident. But i will "STILL" say this, i do not think neural nets will ever work as a end all to trading, the human mind can never be copied where a computer can switch gears as fast and accurate as the mind. The author of this thread and places like tradestation etc might feel different......
Well, that's a relief to me.. I wrote my own ATS which is very reliable.. I don't even know what its doing during the day and don't care. Also, neural nets are very powerful but you have to be very very specific about what you feed into them and what you expect to get back. Good luck, Stephen
I have used neural networks to develop trading systems since the early 1990's , when I developed BrainCel and developed Trading systems for a large investment bank. I have also written a lot about developing Trading Strategies with Neural Networks, Genetic Algorithms, Fuzzy Logic and more over my decade writing for Futures Magazine. I agree with you 100%. In fact I would go a step further you need to be careful what you ask neural networks to do in a system. My research shows you need to ask them in improve a already tradable system. This way if they fail , the system only degrades but does not implode.
There are enough trading opportunities in all markets for both kinds of traders. Some prefer to wait and only deal with the same setups repeatedly, filtering out the rest; others do not mind perceiving trading as a series of unique events/opportunities to be dealt with individually. Most individual traders prefer to trade/perceive mechanically because frankly they only have the quotes and charts to work with -- they don't have the access to resources which would allow them to define each setup beyond numbers/patterns. Compare that to institutions with access to order flow, customer patterns, commercial hedging schedules etc. If you have access to the tools, you use them; otherwise you make do and play with the toys you have. I would rather think that our perception of trading starts more as a science and evolves into an art; in other words, trading as craft. Once we get the mechanics down pat, we move on to more extraordinary things. As Ed Seykota put it: Follow the rules without question. Know when to break the rules.
For a 100% mechanical trading system with no discretion, what would be good performance measures before one trades it? To those good and experienced mechanical traders, is a system with 46% winning ratio with average winner being 2 times bigger than average loser in a daytrading strategy, good enough? Or would you prefer a 65% accuracy with average win/average loss equal to 1? Of cource the former has a slightly higher expectancy. I am now almost 100% mechanical in my trading. But there are many times, my gut feel is right but I still stick to my system. Why? Because of the comfort that my system has been backtested and is robust over many markets, although I only trade 1. Many times, I think I can more accurately "predict" what the market will do than my system, using my reading of the peculiarities of the chart, but I can never quantify these, and also because of the difficulty of having discipline in executing less rule-based discretionary trading, I continue to trade 100% mechanical. These might be holding me back, but then again, if I am still profitable, I must be doing something right, I guess. Would like to hear other's suggestions, especially profitable ones.
loosenup I would venture to say you are describing your "INTUITION" when you see things in the mkt. Am i close?.... Intuition is your brain telling you: I have seen this before and i have within my memory bank good feelings from past experiences of this image now being projected to said GREY matter. This is the neural network of life, your past experiences.
I recommend you study the entire equity curve rather than summary statistics. If you are deciding between 3 systems, slurp their equity curve data into Excel and plot them all on the same graph. If you're down to one system and it's a go/no-go decision, get in touch with your intestines and determine whether you are comfortable with that equity curve. Instead of worrying about "what numbers are GOOD numbers?," consider the question "can I emotionally withstand this?"
I like your good thinking/ experience. However, for automated trading systems, should I say "I'm a skilled discretionary trader with majority of trades using human-assisted automated order/ trade execution systems."? TIA.