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] [53] [ 54 ] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81]
54 Part IV: Bollinger Bands with Indicators If you are correct and the indicator behaves as expected, you may employ the idea. Purist advocates of first principles will argue that there shouldnt be any modification to the original formulation, but in my opinion there is no need to be so hidebound. Test and adjust and optimize carefully, taking care to avoid the usual traps, and you should be fine. Optimization is a topic well beyond the scope of this book, but one cant discuss indicators and systems without touching on the subject. Optimization is littered with pitfalls that can entrap the wary investor. While optimization can be a useful tool, it is often abused, sometimes unknowingly. The product of this abuse is merely a good description of the data rather than a useful tool. The abuse of optimization is another way into the sycophant trap that was discussed above. Optimization is the process of finding the "best" parameter(s) for a given approach. These days, optimization is usually done by computer, but it was done by hand before PCs were common. The simplest and most common optimization is the moving-average crossover. The optimization program starts with some small value for the moving-average length and then calculates all the buys and sells based on crossings of the average, and reports the profitability, number of trades, worst loss, best gain, etc. The process is then repeated for a slightly longer average, and again for a slightly longer average, and so on, until some terminal value is reached. The results of all the runs are tabulated, with various statistics computed that allow the user to see what the most profitable parameter(s) was. The optimization process can quickly get quite complex. For example, lef s examine optimizing a system using Bollinger Bands and one indicator. Lefs say you vary the average length by 2s from 10 to 50 (21 trials) and the indicator period by 2s from 4 to 20 (9 trials)-21 times 9 equals 189 tests. Now vary the width of the bands and the threshold for the indicator, say by just two levels (3 trials for each) and you have 1701 trials-189 times 3 times 3. You can see how it is possible to get mired quickly. Sectioning is one way to avoid the most common optimization pitfall-simply building a good description of the data on hand. Break your histories into several different sections and perform your testing independently on each. For example, if you had 10 years of data from 1990 through 1999, you might optimize in
Chapter 17: Bollinger Bands and Indicators three sections of three years each, using the first year of each section as a run-up period for the indicators and the last two years for optimization, 1990 through 1992, 1992 through 1994, and 1994 through 1996. Then test the results for consistency on the most recent period, 1996 through 1999, again allowing the first year for run-up and testing on the final three years that hadnt been seen in the prior optimization runs. The results from each section should be quite similar; the greater the similarity, the higher the confidence you can have. This is called robustness. Another tack is to break the items you are testing into several different groups, perhaps with different characteristics, e.g., volatile versus stable, growth versus value, small versus large, or low price versus high price. Look for consistency in results. The idea is to assure that you actually have a window onto important analytical information, not just a good description of what worked then, or what works for those stocks. One last test is to see whether the parameters you have chosen are robust: Change your parameters by small, but meaningful, amounts and retest. If you have a robust method, the results of the tests should again be consistent; i.e., if you find 20 periods to be optimal, then 18 and 22 periods should produce similar results. Next well look at volume indicators, and then well consider two methods based on indicators confirming price action in and around the Bollinger Bands. KEY POINTS TO REMEMBER Use indicators to confirm band tags. Volume indicators are preferred. Avoid collinearity. Choose your indicators before the trade. Use prebuilt templates for analysis. If you must optimize, do so carefully.
CHAPTER VOLUME INDICATORS For those of you who want to fine-tune or alter either our methods or your own methods, this section provides the background information youll need to do so effectively. Those of you suffering from math fright may skip the second half of this chapter, but do read at least the next few paragraphs. Volume indicators are the most important group of indicators for the technician. They get right at the heart of the supply-demand equation while introducing an independent variable, volume, into the analytical mix. Underlying all volume indicators to some extent is the concept that volume precedes price. For example, during a base, smart investors are accumulating stock in anticipation of a rally; or in the latter stages of a rally, smart money starts to get out before the top is in. Volume indicators suffer from terrible nomenclature problems; they are rarely referred to by the same name in any two programs.
[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] [53] [ 54 ] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81]
|
Menu
