I'd like to know how many traders actually automate their trading strategy and if any are profitable in the long run letting the system trade on its own. If you could expand on what type of strategy you use, that would be great (gaps, moving average, MACD). I also wanna know how many of you actually use hardcore programming devices such as neural nets and other fuzzy algos and how these influence the decision making process. Thanks a lot.
Me for one. I write my own system and I've thought about fully automating my trades. I'm sure I can become a fully automated program trader. But I outperform my system so better off not automating it. Also, no matter what, I have to stay with the computer even it is automated due to some bad datafeed or some other problems beyond my control. Neurals? Fuzzy? I've tried out a few things but it just won't become better than me. Most of these are based on Fractals which is good in a sense. Still not an expert in this field. I've considered myself as a system trader but I've recently noticed that I'm a discretionary trader who programs. For my trading style. I'm a top and bottom picker that relates them to the trend. I use trendlines for intraday and Gann Angles for daily+ charts. The system I use confirms my analysis from a momentum and volatility standpoint. I also have a few filters in them based on my discretionary style.
We have a fully automated system in place. We've just recently started publishing our results for the purpose of sharing them with other system researchers ( MarketHackers.com ) Check it out if you're curious. As for the specifics to the system, those can't be described, but overall architecture can be. The system is programmed in Java utilizing the TWS IB API. Candidates are selected with a nightly process. Our quote provder is PCQuote MarketSmart. To generalize the quote data process, we utilize the QuoteTracker http interface. This way, if we ever want to change data providers, as long as its supported by QuoteTracker, we don't have to make any quote changes. Neural nets have been explored, but after evaluating several standard TA systems, we concluded the extra complexity wasn't worth the return.
of my own work. I've developped an automated trading system completely modular in C/C++ GTK under Linux. It is interfaced with TWS API (bypassed the annoying TWS window connection socket yeeshhh). The feed will probably come from MyTrack which offers a pretty good API. It's going live in a little less than a week I think. Some of the code will be made available, I believe. All I can say is that it's fast and stable as hell and I'm glad I'm not using Windows for it 'cause it might need to be clustered (yay for Mandrake clustering). I will probably if not certainly use some of the Neural Net / Genetic algorithms libraries available for linux on the web (another advantage of linux). I will make a lite version available for Manual trading so my fellow traders who want to trade under linux can Oh and it's MDI so you can multi monitor it.
"Probably" going to use some libraries, implying that its not implemented (and the system has not been finalized), yet you state it will be available in a week? And with all of this code slapped in together, its still "stable as hell"? Hmmm.... Systems that are stable, especially distributed systems, are not slapped together. Please post the url to your system when available.
are not implemented yet and they won't be for quite some time. They are not part of the core program. You don't really need them in most situations. I want to experiment with them first.
of things. The program itself is not a system, it is a program or interface. The algorithms developped sort of plug into the program. That is probably why you thought the program couldn't go LIVE yet. I have a couple of algorithms already programmed that I have to test. Checked out your site. Impressive results.