back start next

[start] [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [ 43 ] [44] [45] [46] [47] [48] [49] [50] [51] [52]



average profit displays a nearly tenfold range from $.70 for ALA to nearly $7 for CHRP.

Compare this with Figure 13.8, which graphs the same stocks during the same 60-minute breakout period, measuring them by the percentage of days during which a breakout of the 60-minute range occurred on only one side of the intraday range.

Comparing the responses of the same two stocks as before, the percentage of successful breakouts for ALA is 70 percent while CHKP is 75.82 percent. Although the two issues seem roughly equal when compared as a function of successful breakouts, the profitability of trading CHKP is significantly higher than would be indicated for ALA when one considers the relative profitability of trades generated by each issue.

Once again, here is statistical evidence of the wide variety of responses of different issues to the same parameter, enforcing again the necessity of customizing your strategy to each issue to be traded rather than expecting all stocks to be traded well by the same system.


90.00 »


£ 50.00


0.00 -

"* 5 5 £ s < 9 < < * *

1 5 % i 1 I 3 1 £

3 i 8 3 g ° 1 \


: to \

: t 5

\ 8 s g s j

Figure 13.8 Successful one-sided breakouts of tbe 60-minute time frame, expressed as a percentage ofthetotaldaystested, are grapbed for a selected number of issues. Note tbe significant variance in response between issues.


1. The Directional Day Filter has another use: Its data can be used to fabricate a breakout trading strategy.

2. Careful observation of historical data is important in setting the parameters for the breakout.

3. Each stock issue will respond somewhat differently to this breakout scenario.

4. Data generated from a significant database validates the theory that one side of most daily ranges is formed rather early in the session.


So far we have concentrated on the use of several indicator combinations to develop a functional day trading system. The effort thus far has been primarily directed toward those persons who wish to trade from a discretionary standpoint, taking their clues from machine-generated data, and graphic plots and, in general, experienced chart observation.

Although totally automatic, computerized trading systems are not the focus of this book, a brief overview of the process is included here for those who will want to take system design and implementation to the next level.

Near the conclusion of Chapter 12 I discussed the difficulties that traders have keeping track of the various combinations of oscillator indicators, support and resistance formations, and information gleaned from the use of the Directional Day Filter. Use of an automated trading system can significantly simplify trading such a strategy, as the programmed system can consider all of the above when generating a single trading signal.

I often relate to traders that it is necessary to use not only the computer on your desk but also the one between your ears. It will be a very long time before we are able to duplicate a human mind on a

silicon chip. Although we are able to automate many of the mathematical procedures that lead to trade generation, traders must always leave themselves some room for input.

The previous strategy places considerable dependence on human intervention as computerized indicators generate the various stochastic, RSI, and Percent R plots from which we are able to make our trading decisions. Automated systems actually take into consideration all of these factors and generate the trading signal at the proper time and price.

These systems by their very nature must have strictly defined parameters from which to operate. While this feature assumes some of the decision process from the trader, the advantage of such a system is that it provides significantly more options from a testing standpoint. These systems, when applied to historical data, will provide a picture of profits and losses that would have occurred had the system actually been trading over this time period. Various combinations of system inputs can be tested to give the trader multiple options for either trading or additional testing.

The testing of trading systems can be a significant advantage in that the trader is able to assess the effectiveness of a system and make adjustments prior to placing actual trading capital at risk in the marketplace. Again, as mentioned previously, trader confidence is a big part of eventual trading success. Having a system that tests well against back data certainly can increase ones confidence level.

However, automated system testing can also be a significant detriment to a trader. Since todays software is capable of testing multiple system settings against historical data, the possibility exists that these parameters can be tested to the level that they report a fantastic profit picture over past data. This type of testing, also referred to as curve fitting, gives an unrealistic picture of historical performance. The system parameters are so closely fit to the historical data to which they were applied that the system has virtually no chance of repeating this performance in real time. Many traders have learned this one the hard way as their super system gradually crumbles before their eyes in real trading.

These problems can be avoided by the use of system testing applied as an "out of sample" routine. This procedure involves the testing of a given amount of data, six months for example, and then applying the resulting system parameters to the next six months of

data. The results of the application to the next set of data simulates real-time trading to some degree. In this fashion the system parameters are not curve fit to the testing data and are therefore more realistic. Although this routine is complex and beyond the scope of this book, be aware that there are productive uses of system testing that can be a definite benefit to the design of a successful system.

Perhaps the best use of an automated system for the purposes put forward in this book is to use them as a method by which to determine the most effective settings for dual oscillator indicators. By applying these indicators as systems one is able to get an idea of which settings are the most effective for the particular market and time frame of interest.

In this chapter I am going to take you through this next step, briefly exposing you to the world of computerized system development. This type of system structure also requires the use of the computer between your ears, but more on the front end, where the strategies are developed and tested, and less on the back end when the signals are actually placed into the market. The next few paragraphs cover the thought process one could go through when designing a system to test the breakout strategy that revolves around the intraday range as determined fairly early in the day by the Directional Day Filter.

As you will recall from the previous chapter, one of the major steps in formulating a stock trading system is actually selecting the stocks to trade. There are many routines available for this task. In this instance I will make use of the statistical analysis performed previously to demonstrate the differences in reaction of the portfolio to given system parameters. Table 14.1 reflects the relative rankings of the top 15 stocks from the database, rated by the average maxi-niu-iii profit attainable by each from a long breakout.

This list represents the issues that look as though they should respond well to the system in general. By applying a computerized trading system to each of these issues one is able to make several observations relative to the performance of the system. You will recall from previous discussions that the response of individual issues to a particular system will vary significantly. By applying the same system to each of the 15 issues in the list, one is able to define the system further, altering it slightly in such a fashion that it can be expected to respond in an optimal manner to each issue according to the specific

[start] [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [ 43 ] [44] [45] [46] [47] [48] [49] [50] [51] [52]