I've played around with automated trading for several years with limited success. I can get my different ideas to work with Excel receiving data and sending trade signals to Interactive Brokers. The problem is that whenever I test something for 100+ trades, the result is always close to breaking even. Does anyone have any idea what method of trading would give me the best chance for success? Which equity works best with automated trading? I've tried all different methods, moving average crosses, breakouts, buy at the lower band, sell at the upper band etc. etc.
Hi, what platform are you using? I recently got into automating and have successfully made several intraday strategies (im primarily a swing trader). Here are some tips: o if you go to really short time frames, you get more random movements than longer ones. However, short-timeframe work better for pure momentum systems. o What are your goals? Maybe they are unrealistic. I find that my win rate goes down for intraday since there is substantially more noise/chop. Maybe its worth capitalizing on. o If you are testing on futures, for example, you might only want to include the times they are actively traded. The overnight sessions when they go sideways can give 'false' signals (depending on your setup )which dramatically can dramatic impact the results. (Ie you need some kinda volatility filter). o I'm relatively new to intraday development, but so far I've found daily setups to be far more accurate/worthwhile. You might want to check the art collins threads for some 'edge' ideas. His published work on the threads here so substaintial profit buying open selling close (I think) not including comms + slippage.
Sam Are your ideas based on indicators? I am interested in automated trading but know very little about it but I have set up a couple of things in Quotetracker using their paintbars which give me long and short signals so I assume it would not be a big deal to transmit these signals to IB but Im having the same problem that you are in that the inherent lag in indicators results in flat results. If I could figure out how to get my signals from price bars rather than indicators I think the results would be much better but this is over my head. Does your system use indicators?
Right now I'm using Excel to get data from and send orders to Interactive Brokers. I have it set up with Sierracharts to retrieve data and Bracket Trader to send orders. Eventually I'd like to simplify things and just use Excel and Interactive Brokers without resorting to the intermediate programs. I'm using short time frames, five minute bars right now. That makes testing methods quicker and gives me a lot more data to work with. I suppose I would probably have more success looking at longer time frames. My goal is to be able to make any kind of program that is profitable. It doesn't matter what time frame or equity, as long as it can make more money than just breaking even. An example of the latest thing I was trying is as follows: Set up a five minute bar chart of ER2 futures. Set up a 5 bar moving average least squares of the high for the upper band and one of the low for the lower band. Set up a long term, 10 bar exponential moving average. Calculate the slope of the long term average (average now-average 5 minutes ago) If the slope average is greater than -0.03 then buy on the lower band and exit on the upper band. If the slope average is less than 0.03 then sell on the upper band and exit on the lower band. It is relatively difficult and it takes a lot of work to learn how to automate a method and then get it to work. What I'm finding now is that it must be even harder and even more work yet to find a method that actually makes a profit.
the main problem is that you have to test as many trades as possible. like 20,000 to 100,000 trades per method then you will know better about what a system will do. mb
personally, i prefer a tick or volume charts for auto trading. it allows one to use (if they choose), 24hr time frames. minute charts won't do this as well.
Agree. The conditions you've seen for the past few years are unlikely to repeat in the same fashion for the next 10 or more. Although it's also unlikely that any prediction I make will be accurate, it's difficult to envision the convergence of the various economic conditions that we've just experienced, repeating again any time soon ... R
I agree but i think its not an absolute measure. BUT its important to test across a statistically significant number of trades. My weekly system generates about 40 trades a year. If I wanted to test it over 20,000 trades, i would have to go back to the 1500s LOL
yea well if thats what it takes cause even a few thousand trades can have random results which you think are solid till the data changes. they make random data generators for guys like you who have weekly systems - there was one named metrodome or something like that. mb i may have a copy i could send you.