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] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95] [96] [97] [98] [99] [100] [101] [102] [103] [104] [105] [106] [107] [108] [109] [110] [111] [112] [113] [114] [115] [116] [117] [118] [119] [120] [121] [122] [123] [124] [125] [126] [127] [128] [129] [130] [131] [132] [133] [134] [135] [136] [137] [138] [139] [140] [141] [142] [143] [144] [145] [146] [147] [148] [149] [150] [151] [152] [153] [154] [155] [156] [157] [158] [159] [160] [161] [162] [163] [164] [165] [166] [167] [168] [169] [170] [171] [172] [173] [174] [175] [176] [177] [178] [179] [180] [181] [182] [183] [184] [185] [186] [187] [188] [189] [190] [191] [192] [193] [194] [195] [196] [197] [198] [199] [200] [201] [202] [203] [204] [205]
32 I S Sits iSsssiiliiiilliiilHiiJililll sssssililiiiililliiiillllfllllll -iiMIIIšilllšl!§liliiišlšll .iliiaiJIillHisiiiiSliillKliill iililiiKlllilHilllllHillil »lilliiillllliiilliilš§SIIš liiiiiililliiillfliliiliilii FIGURES-10 Trading sj-aem using two moving averages. Close oui long tvhen Donchians 5- and 20-Day Moving Average System Tlie combined use of two moving averages has been popular among professional advisors. Of these, the most well-known is Donchians 5- and 20-Day Moving Average, a method claiming one of the longest recorded operational results, beginning January 1, 1961. There is no explanation for the selection of these two values, but in 1961 when moving averages were considered state-of-the-art, they were a reasonable choice. The selections can be justified because of their close relationship to the number of trading dajs in a week and a month; even now, similar calendar periods might add some desirable features to a sjstem. Donchians idea is to use a volatility-peneU-ation criterion relative to the 20-dsy moving average, but with some added complication. The current penetration must not only cross the 20-dsj moving average but also exceed any previous 1-dsy pendration of a closing price by at least one volatility measure. The 5-dsy moving average serves as a liquidation criterion (along with others) and is also modified by prior penetration and volatility. These features tend to make Donchians volatility measurement self-adjusting. Even if selected poorly, the new penetration must exceed prior breakouts; without thorough testing, it may be difficult to
determine wliich rule has more significance. To maintain a human element, Donchian requires execution of certain orders to be delayed a day if the signals occurred on specific weekdaj-s or before a holiday. The combination of different factors is generally the result of refinement over years of actual operation, but neither complexity nor sophistication guarantees success; only the results will tell. More Than Two Moving Averages If two moving averages improve trading, it should follow that three or more are even better-but it doesnt. With the use of two moving averages, there is a main trend identifier and a timing device; the addition of one or more averages or indicators must fulfill a distinct pu ose. When multiple moving averages must agree on the same signal, there is less TicLardD Doncluan "Donchians 5-and 20-Dav M-viu? Aver.wes " Cjuniodities . . : ceniber 1< 11 chance of getting a trading signal and less time holding the trade. The results of moving average sjstems using two and three moving averages are covered in Chafer 21. COMPREHENSnT STUDIES Because computerized testing programs have made it easj for anyone to test any number of trends in combination, there have been very few comprehensive shidies published within the past 10 years. Nevertheless, there is a great deal to learn from putting the results of various sjstems and markets side by side. Earlier in this chapter there is an informative comparison of six major frending sjstems; others shidies have been included in Chapter 21 ("Testing"). Among them are sjstems similar to the single and double moving average methods just discussed, as well as the selection of a third frend for entry timing. An important objective of the testing process is the longevity of the parameters selected: this discussion can also be found in Chafer 2 I SELECTING THE RIGHT MOVING AAmAGE Up to now, the selection of the right moving average, the one that will work in the future, has only been discussed in general terms. The best moving average speed for a hedger may not be the same as for a speculator. For , a hedger may be marketing cattle each month and may need to choose a time at which to preprice his or her product. A 3-day moving average might generate five sell signs in I month, each the result of a 2day price move-an ineffective tool for the hedger. The hedger needs an average of one sell signal per month to be a useful timing tool; the speculator has no such concems. For the speculator, the right moving average speed is the one that produces the best performance profile. This profile could be simply maximum profits, or it could be a more complex combination of profits, equity variation, and use of margin. In Chapter 21. the computer is used to find the combination of speed, stop-loss, and other rules that besl satisfies a preset criterion; a computer, however, is not alwajs necessarj. Computer testing of a trend sjstem or other trading strategies sometimes leads to solutions that are highly fitted. The computer may find that a 3-day moving average was more profitable with lower rid; than a 20-day frend which was only slightly worse in performance. Our logic tells us that the results of the 3-day sjstem will be more difficult to attain in real trading, because execution costs in fast markets with many frades may he higher than expected A slower selection is more conservative and more likely to retum the expected results. I orever, the trader must be the judge of whether the selection is reasonable. A grain trader knows that the seasonal price pattem forms a complete cycle each year. At best we can expect one long upward move followed by a shorter, faster decline. Not all trend speeds can capture the profits in this pattem. For example, if the upfrend lasts for six months, a 6-month moving average wont see any of it; therefore, you will need a moving average not much longer than 1/4 of the length of the trend and perhaps shorter. If mass testing of moving averages using an optimizer results in a best moving average period of 6 months, that choice must be understood as a failure of the sjstem to find any shorter period that worked. Selecting theTrend Speed Observation of a price chart will show the trends of the market. The trends that two traders see will often be different. Some fraders immediately focus on long-term price trends: PeirvJ rjubuan,"JJ"Viu2Aver?E in Tra.liu? Tactics ALivrf-ckFutures Aiitliol-y, Toddbfbii (ed) ( . . . , ISBfii
Traie | MwelWeek | Pmfhabk Weeks | Total Profit | | 1.30 | | 10.4 | | i.2S | | | | 1.40 | | | | 1.70 | | | Total | | | 30-6 |
Naturally, there will be periods of uncertain direction and small losing frades with this or any price series; however, capturing these four price trends win go a long way toward profitability MOVING AA-ERAGE SEQUENCES: SIGNAL PROGRESSION Consider the case in which you have selected a 20-day moving average to trade. You enter the day with a long bond position and you get a sell signal. However, you dont know because you are not looking, but the 19-day and 2 1 -day moving averages did not get sell signals. This means that the day that was dropped off the calculation 20 dajs ago caused a slight shift not seen by the other neighboring frends. Is this important? Yes A moving average is simply a consensus of direction. It is an approximation of values intended to put you on the right side of the market more often than not. It is most fallible at the time prices are changing direction or going sidewajs. Any information that clears up the problem is helpful. For any moving average sjstem, we would like to see a steat progression of frend change from the short term to the long term. This is seen in the following illustration: Example El. MovingAverage Period in Days 1 2 3 4 5 6 7 6 9 1 11 12 13 14 15 16 17 18 19 20 Zl 2Z 23 24 25 Trend uiiuLuuuiiuuuuuiiULuuuclddddd FIGURES-11 Selected trends and reversals. others see much shorter movements. To find your own best moving average speed without the use of a computer, look at a price chart and mark the beginning and end of eadi price move that you would like to capture. These trends may occur every few dajs, or only three or four times eadi year. In this exercise, the ma}or trends were selected from the chart for August 1985 pork bellies (Figure 5-11). Use the fad that a moving average of n-day speed will be neutral over a period of n dajs in which the price retums to the same level. (Remember that a 1-year moving average is used to remove the annual seasonal efferts.) If the daily price changes are nearly equal, one leg of this move (the trends identified in Figure 5-5) win occur in n/2 dajs. To be cautious, select moving average intervals slightly longer than two times the worst retracement within the frend to prevent being stopped-out of the trend. When calculating profits generated by a moving average, pan of the trend movement is lost at the beginning and end of the frade. This part is equivalent to the price change over V, of the moving average period, because il requires that much time for the moving average to break even, or become neufral with respect to the price move. Keeping these two points in mind, the best trend speed can he selected from presenting the largest moves and largest reversals, as shown in Table 5-6 The longest period needed to offset the largest reversal is 3.1 weeks (eadi week is 5 dajs), and ittakes a 16day moving average to neutralize the largest move. A 16-day moving average usually requires 8 dajs to break even. Allowing 2 weeks (10 dajs) of price movement to readi the break-even point, the following profitability remains:
[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] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95] [96] [97] [98] [99] [100] [101] [102] [103] [104] [105] [106] [107] [108] [109] [110] [111] [112] [113] [114] [115] [116] [117] [118] [119] [120] [121] [122] [123] [124] [125] [126] [127] [128] [129] [130] [131] [132] [133] [134] [135] [136] [137] [138] [139] [140] [141] [142] [143] [144] [145] [146] [147] [148] [149] [150] [151] [152] [153] [154] [155] [156] [157] [158] [159] [160] [161] [162] [163] [164] [165] [166] [167] [168] [169] [170] [171] [172] [173] [174] [175] [176] [177] [178] [179] [180] [181] [182] [183] [184] [185] [186] [187] [188] [189] [190] [191] [192] [193] [194] [195] [196] [197] [198] [199] [200] [201] [202] [203] [204] [205]
|