Technical Analysis- Indicators & Systems Message Board

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Strategies & Market Trends : Technical Analysis- Indicators & Systems

Public Reply Prvt Reply Mark as Last Read File

Previous 10 Next 10 Previous Next

To: Bruce A. Bowman who wrote (1382)	6/10/1997 9:48:00 AM
From: David Russell Coburn II	of 3325

Hello Bruce, It's just when you get to writting about these types of things you start getting so many thoughts and directions and ideas that it's impossible to relate them all effectively. But thank you very much for the compliment. "1- be sure to correlate your test results with your actual trading results. If they differ by a significant amount, you aren't trading the way you tested and you need to fix something: either the test or your trading." I agree 100% percent. "2- this is the hard one because it demands a lot of data which often pushes you into stocks you might not otherwise trade. Ideally you would like to test at least 2 separate ranges in your data that don't overlap. Look at the results of the 2 separate test summaries and validate that over enough time and different stock symbols that you see results that are consistent. This attempts to resolve the effects of inadvertantly curve-fitted tests." Not quite sure what you mean by this one. By ranges do you mean time frames or price? Probably time frames I think. I'm not clear on your logic or why deviding the larger time frames into 2 or 3 non overlapping time frames and testing them individually will keep you from curve-fitting the results. Why would you not get the same broad information from just testing many stocks in many different markets? "This isn't something that you addressed, but "curve-fitting" gets a lot of bad press. Ideally you'd like to trade a system that's a curve-fitted system. The problem is that the curve is constantly changing. A system that can identify the characteristics of the curve and select the right indicator(s) and parameters for the indicator is resting on the shelf directly below the Holy Grail in that far away place, just waiting to be found." I havn't tested these ideas but it sounds plausable that a system that is curve fitted to a stock or a group of stocks that have the same types of variables could be very useful to trade that particular stock or group of stocks. As an example, if you had say over a small time frame (6months to 1 year) a group of electronic video chip makers with floats under 15Mil in the price range of $10 to $20 with similar volumes etc., etc.. They might have similar trading frequencies and or curve actions. In this case there probably is an optimum system for this group that doesn't work nearly as well on other stocks. The problem would be the time involved with developing the systems and monitoring the changes of the curves so your system doesn't become un-fitted. I think a more realistic method might be to be not quite so fitted. First have different scans that look for particular types of stocks and then a system that you trust and use for that type of scanned stock. There might be an advantage in the diversification of having the different types of stocks covered. 1)slow smooth trending stocks 2)high volatility stocks 3)periodic at different frequency stocks 4)high price, large float stocks 5)low price, small float stocks The above is off the cuff and I am not advocating it but it might be an interesting thing to test to see if you get an advantage either in trading a particular type of stock or from being more diversified. Thanks for sharing your ideas. David PS. In response to your next post on historical splits and its' effects on testing, you are correct in that this is another thing about system testing that will cause my "what type of profits can you expect" idea to be inaccurate. I'm not quite sure how to get around this except that if I wanted to take the time to look at the history of the particular stocks splitting over the test time frame I could adjust the system buy criteria. Your point is well taken. Again, testing accurately is very tedious and takes alot of time. But in the long run I have confidence it is well worth the effort. The subtleties can eat your lunch. Your first point of correlating your actual trade p/l to your test numbers is very important.

Report TOU Violation

Share This Post

Public Reply Prvt Reply Mark as Last Read File

Previous 10 Next 10 Previous Next