[amswak]

Here is my experience with automated trading:

I first started daytrading a little over a year ago. I failed miserably. I had about one good month, then the fact that I didn't really know what I was doing caught up with me...fast.

That was when I decided to start coding automated strategies. I am an engineer and the code came easy, so I thought it would be a cinch. Again, the fact that I still didn't know anything (other than now being more fluent in EasyLanguage) caught up with me.

Four months ago I stepped back and went the discretionary route. My favorite indicators are channels (price and bollinger), John Ehler's smoothed adaptive momentum, and an adaptive RSI. I have had great success.

Last month I started trying to build automated strategies again using my favorite indication combos and, what do you know, they are looking promising. They automated strategies are still in simulated mode, but at least they are showing promising real-time numbers.

I heard an excellent analog to this recently: You may be able to parallel park your car easily yourself, but telling a computer to do it for you is a significant undertaking. And in the event that you can't parallel park on your own in the first place, it will be near-impossible for you to make a computer program that does it for you.

Also, if you dont believe automated trading works, check this out: HFT article
42% of the market!!! Now I know HFT is different from "our" kind of automated trading, but it is still automated trading...

I think you get the point. It is possible. Difficult. But Possible.

Happy Holidays!

Hunter

[Uli Schmuli]

Hunter,

Do you have a spy cam in my office? My experience exactly, right down to the being an engineer part ;-)

Uli

[SangreDeToro]

absolutely possible - i went down this route because a trade plan must be an algorithm and if its an algorithm it can be coded.

[maxima]

Originally Posted by FulcrumTrader »

Check out TradeVec.....MANY sophisticated automated traders are starting to use this platform.

I tried to look into it again but is still in beta... Do you know when are they going to release the product?

[Szymon]

Originally Posted by cowcool »

I am yearning for a 100 % mechanical trading system that is profitable?. Is this possible?

Thx!

Yes this is possible, a lot of hard work but once you understand how the market works then you can design a very good mechanical system that gives excellent results with profits.

Here is a sample of a system at work using a dummy account. Is uses minimum 20 lots but can be changed to 2,1, etc...

Good luck

Regards

Simon

[andy2m]

I also think it's possible, actually I hope so :-) I've written my own simulator/executor of mechanical strategies, now I'm able to backtest (and optimize) strategies and once I will have found one I can put it into the executor to trade it fully automated.

I'm in the process so I can't say if this adventure will have a good end. What I can say is that it's possible to build up a machine that is able to trade even complex ideas. After finishing my software I was sometimes looking at the executor what he did and was amazed how he put in the orders, moved the stops and so on, nice to watch. Of course that should be possible, it's just an engineering task but my question before starting was: will such a system be technically stable and safe enough (so that I can put real money on it). My answer is: yes. When I trade real money then a 2nd computer runs in my parents house to observe the trading computer and in case of anything it executes going out of the The technical basis is done.

I think there are many advantages in technical trading, you can do other things, can trade many markets simultaneously, even at night, fast in executing (scalping possible) . It was a lot of work writing my own machine but before I checked any other system I could get (2 years ago) and I didn't find any good one for me. I've done some extended trys using Tradestation but I was always disappointed that the reality was not as good as the simulation. I believe one reason is the accuracy of the data. My own system uses 5sec bars and additionally tick, volume and order sizes and now it's accurate enough, means when I run a real trading day again in the simulator it's about 95% correct - that's enough to be able to find working systems in the sim.

The big point actually is to find system ideas. I say to myself: if I have found only one that works there will be many more. "Works" is definded as constantly making money, steady upslope capital curve etc. at least over some weeks/months. So far I've only dealed with relatively simple strategies without "self learning" stuff. This can work very good for some days but then "suddenly" there comes a "painful day" where it gives away a good amount of the money earned before. E.g. I have a strategy that is mostly successful in fading the 2nd breakout: on normal days I get more gain than loss, sometimes win sometimes loose but with good money management I win. But then it happens that there comes a real trend day, the market going in one direction steadily - and that's pure poison for this system, it looses the whole day. I made the same experience using other trading ideas and the result is: with "simple ideas" it's possible to program a small gain over long time, not more. The market (I actually trade only Eur/Usd) is very clever finding new ways in behaviour even if in the end it's the same again and again.

Things like Ross Hooks are easy to gauge and I thought "yes that must work", after programming the machine tells it's own version: yes it works often, but there are also many times where it doesn't work. I looked back at the chart and had to agree, the machine (= the pure technical idea) was right, I had overseen some non working RH.

So this is my contribution to the thread starter's question. I still believe it's possible but one must find a working strategy. This has to be easy, adaptive and clever at the same time :-) I'm actually working on integrating a Neural Network and other "learning stuff" into my system and I believe I will overcome the drawbacks mentioned with it. I also do some statistical research on daily behavior (like Taylor swings) to be prepared for special types of days. I don't think I will able to avoid losses but I want to avoid the "painful days" as I have them now. The market repeats itself and in the same moment it changes it's behaviour constantly, and the changes repeat themselfes. I believe a mechanical system must be able to deal with this fact.

Happy trading!
Andy

[maxima]

So Andy tell us.. You wrote the whole backtesting system from the scratch in just two years? Is that right? You must be talented young fella.. What programming language do you use?

[andy2m]

Yes I did. I'm a professional software developer so I know my tools. But I must admit it's my 3rd attempt, before I have written two systems in Excel which didn't really help me, but there I learned about markets, indicators, data etc. I also had tried to trade manually but have found that's not my thing.

Evolution: First I wrote a small trading robot dealing with the API of my broker, written in Visual Basic 6, it already had the capability to attach different strategies and could send orders, react on fills etc. Then I decided I need a simulator so I made it work with collected data from the past. But that was too slow (it needed 3mins to simulate one day using 5sec bars), I ported the engine part to C++ (now only 2secs for one day). This made it possible to add a parameter system, means parameters are read from a database and performance results are stored back, so I can do brute force optimizing. The last step was to synchronize the source codes of both simulator and trading robot so that now I can use the same strategy code in both.

Of course this is easier written than done :-) It was a lot of work and I wouldn't start that again. I had flexible jobs at that time so it was possible.