![]() |
![]() |
|
|
Pattern Recognition & Trading Decisions - FAQsFrequently asked questions Using the programsQ1. How do I use the programs on my own data? A1. Two feature articles have been prepared to try to address this point. They can be downloaded from the section of this site dealing with feature articles for the book. Q2. How do I use Indicators.exe to find a stop level? A2. Like trading styles, stops are very much a matter of individual preference. If you start up Indicators.exe with your own .cmx data, click on Indicators ? Reg Filtered Line and enter “1” (for one line), register it, then enter a length (typically 13) a regularization (probably 0) and a stop width (say 1.5). A graphic display will appear of a stop-and- reverse trading strategy, with stops shown as dotted lines. It is then a matter of playing with lengths and stop width parameters in the program until you see a stop policy that makes the most sense for your style of trading. Bear in mind that the red line plotted over price is the datum from which the stop will be hung or built. If the stop and reverse system’s current position is in the opposite sense to yours, then to find your stop level you will need to project the distance between the (central) filtered line and dotted line over to the opposite side of the central line.
RegularizationQ3. What is Regularization? A3. In technical analysis, there is usually a dilemma between judging if a price movement is simply a random fluctuation or the beginning of a major move. Some indicators (e.g. momentum) tend to be intrinsically noisy. Regularization offers a technique for calculating indicators in a way that penalizes curvature; which is particularly useful for making sense out of noisy indicators without as much lag as would be generated using the alternative technique of increasing the indicator’s length. John Ehlers’ book “Cybernetic Analysis for Stocks and Futures” shows that some forms of regularization are equivalent to a regular filter (two-pole Butterworth type) used by electrical engineers. There is a wealth of literature on regularization in the field of data modelling, but this will have to do for now. Q4. How do I know what lengths and regularizations to use for the indicators? A4. Unfortunately this depends on a number of factors, the most important of which is probably style of trading. Someone who trades frequently will usually use short indicators and vice versa for the longer-term trader. Regularization is usually introduced to be the minimum necessary to damp down fluctuations sufficiently for a user to be able to make sense of an indicator. The programs are intended for manual optimization, where a range of indicators with different regularizations can be displayed simultaneously for users to see what combinations make most sense for their goals. Q5. Can you say more on what values should I use for regularization? A5. The answer to this question is to start small (say 0.5) and increase it gradually (say in increments of 0.5) until the indicator seems over-smoothed for your style of trading, at which point back up until you find a level you like. It is important to note that when regularization is excessive, the formula used to apply in indicator calculations becomes unstable and the results silly. This can be summarized as: use regularization with care, building up from a low level; and be mindful of its potential for instability at high levels which could produce idiotic results.
Classifying marketsQ6. How can I classify markets? A6. Once again, there is a problem of the timescale of an exploitation strategy. A day trader would classify markets very differently to a long-term investor. The recommended way with Indicators.exe is to experiment with lengths and thresholds for the Trend2Noise indicator (as shown in figure 15.4 in the book) to establish thresholds for bull, bear and sideways markets that make sense for your trading. Unlike simple momentum that cannot distinguish between choppy and steady price movements, Trend2Noise can, and is therefore useful for market classification. The general idea in trading is to avoid entering a new position in a sideways market and either hold or exit any existing position in such a market.
Returns DataQ7. How do I get Returns data for the portfolio program? A7. A feature article explains how to do this for a buy and hold strategy. For trading systems, it is much more complicated and requires regular valuations of positions with associated returns on capital needed to support them. Subscription web sites are beginning to appear that can help with these calculations, but they tend to be under-financed and unable to build up the client bases needed to sustain their operations. Q8. How do I decide on a level of return to specify in the portfolio program? A8. This judgement depends on the shape and nature of the efficient frontier. If, for example, a high level of return could only be achieved by investing in only a few stocks of a similar type (remember the dot com boom?) then the level should be lowered to include uncorrelated instruments of a different type. If, on the other hand, a high level of return is achievable with uncorrelated instruments in different areas of economic activity, then the associated portfolio could be acceptable. The success anyone is likely to have with portfolio software depends on the extent of the trawl for opportunities, the quality of exploitation strategies and the diversity that can be built into design of a portfolio. |
||||||||||
![]() |
© 2005-2007, Technical Forecasts Ltd. | Privacy Statement |
website by DesignBase |