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]


171

. Moving average with percent band (MAB) b. Dual moving average (DMC)

4. Sjstems with trailing stops

a. Parabolic Time/Price Sjstem (PAR) b. Alexanders Filter Rule (ALX) c. Combined Directional Movement and Parabolic Time/Price System (DRP)

The study used three measures to test sjstem similarity:

1. The percentage of the time that sjstems are on the same side of the maifcet (long or short)

2. The percentage of buy or sell signals that occurred on the same day, or within a few days of one another

3. The correlation of aggregate monthly portfolio retums

The results show a significant positive correlation in the sjstem profitability (Table 22-5). However, there is no pattem that shows that one particular tjpe of sjstem is notably more correlated than others. The Parabolic and Directional Parabolic sjstems are most similar because one is based on the other.

Only four of the sjstems were significantly profitable: CHL, MIL DRP, and DMC. The correlations of those sjstems do not seem any more significant than others that were not profitable. In Table 22-6, the percentage of trades that occur on the same day is very low. A more informative comparison is seen in Table 22-7, which tallies the percentage of dajs that each sjstem held the same position. This last table is a practical way of finding the performance correlations.

TABLE22-5 Correlations of Aggregate Monthly Retums fi-om 12 Systems

TfodinESyTlem

Syaem CHI PAR UNQ DRP

DRM 0.71 0.61

RNQ 0-70 041 0,70

DRP 0.65 0.81 0,79 0-S7

Mil 0,7S 0,54 0.67 0.73 0.67

LSO 0.59 0,43 0,S3 0,54 0.70

REF 0.55 0.37 0.57 0.66 0.52 0,54 0,60

DMC 0.71 0,41 0.68 0.78 0J5 0.74 0.58 0,57

DRI 0.71 0.42 0.S9 0.77 OSS 0.70 0,59 0.66 0.64

MAB 0-71 035 0.75 0.74 0.6 0.6 0-6J 0-60 D.7B 0.72

ALX 0.58 0.55 0-62 0J5 0.57 0.57 0.58 0.51 0.52 0.56

It. iso i-WiMr drtw, «ii wchrato. iltosot (n

TABLE22-6 Percent of Trades That Occur on the Same Day

S/ram CHL PAR DRM RHQ DRP mi iMJ REF DMC DRI MAB

PAR 19*

DRM 25*" 21**

RNQ 21" 9 15

DRP 30" 93" 48** 10

iso s-vfttlib-dr w. #11m. chrjgo, il i

The study concludes that computer-based sjstems trade on the same day significantly more often than would randomly



be expected, but the actual percentage of trades that occur on the same day is small. These sjstems have the potential to

move martlet prices, but the trades do not often occur at the same time. One must be concerned, however, if trend-following sjstems are holding the same position 15° of the time, they must compete with each other when they exit, even if that occurs over a period of a few days. Trading signaks that occur -at the- exact same time- May not be

necessarj

Correlation When You Dont Need It

There is often a theoretical answer and a practical one to the same question. There is no doubt that a variety of trend-following sjstems generate trading signals at different price

TABLE22-7 Percent of Trading Dajs Sjstems Hold the Same Positions

rroding S/steni

Sywem

DRI MAB

RNQ DRP Mil

75 68 82

83 «9

73 75

59 75

63 58

Dn«.*,IISO.O,laeo,IL6060«(No«mb.r IW7).

levels, some nearer to each other and some farther away. During active, high-volume trading periods, a few of these sjstems can give buy signals without causing prices to move and other trending sjstems to be triggered. During sessions with light volume, it is more likely that a sjstem traded with a large position will cause piggj-backing by others, resulting in a price spike that disappears once all the sjstems have kicked in.

The worst case is a price shock. Although not caused by sjstems, all trend sjstems not holding a position in the direction of the shock will get signals to enter at the same time, the result of a large price jump The correlations mean very litde and only diversification using different trading philosophies, rather than time periods, would help. Although relatively infrequent, large price shocks can be fetal. When designing your trading strategj, keep the practical issues of diversification and rid; management alwajs in sight.



23 Risk Control

A trading sjstem alone will not assure success without proper rid; control beginning with each trade and continuing until a portfolio of different trading methods is created. Sjstems have losing streaks that will ruin any investor with inadequate resources and poor timing; a speculator must decide the initial capitalization, the markets to trade, and when to increase or decrease leverage. There are risks that can be controlled or reduced, called sjstematic risk, and another called maiket rid;, that can take the tom] of a price shock and can never be eliminated.

This chapter tries to cover a broad range of topics relating to risk, including individual trade rid;, leverage, portfolio diversification and allocation, price shocks, and catastrophic rid;. It is not possible to say that one is more important than another, in a specific situation, any one of the areas discussed may be the answer to preventing subdantial loss. The first part of this chter discusses cspitalization and shows why man,.traders m-ill be successful for months and then lose eveiyHiing in only a few clays. It will explain the choices in leveraging and offer alternatives of less rid;. The lad section analyzes when a sjstem is performing property and when it is not living up to its expectations.

Rid; control begins with a trading philosophy the most common is called conservation of cspital. it is the assurance that the investor has been given the most opportunities for success, which usually translates into keeping losses small. This is often accomplished by allowing only small losses per trade or using a frendfollowing sjdem. Once a trend position is edablished, it is held as long as prices continue in the direction of the position it is closed out when the trend changes. The resulting performance profile is one of more frequent small losses and fewer large profits.

RISK AVERSION

Daniel Bemoulli, a famous mathematician, proposed a theory of utility in 1-38 that distinguished between price and value, where price is equal for everyone but value (utility) depends on the individual making the estimate and their circumstances.

Bemoullis spproach defined a concept of diminishing marginal utility, shown in Figure 23-1, which indicates that as wealth becomes greater, then the preference for more wealth diminishes. In the lower left part of the chart, where the investor has a Small net worth, the likelihood of accepting rid; is much higher, although the magnitude of the rid; is still small. \ien rid; becomes greater, even proportional to rex., ard. all ins estors become cautious. Bemoullis graph shows the curve beginning at zero and mov ing up and to the right in a perfect 1/4 circle, ending horizontally, where risk is no longer attractive. This implies that people are rid;-averse. Most people are not interested in an ev en chance of gaining or losing an equal amount. Other theories that have been proposed are that tile maiket maximizes the amount of money lost, and the maiket maximizes the n u mber of losing participants. All of these concq)ts sppear to be true and are very significant in developing an understanding of how the maiket functions.

PeterL Een,.-tein,ed The P.table, IJEA in Inve.-huent G"lm X\iley & Sons, 1 5,p

FIGURE 23-1 Changes in utility versus changes in wealth

/] Decrees Increa.

/ inWeatth ,nwt.e

Initial Wealth

5, -95Pf



[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]