natural rid; profile. A $100 round-turn chaise was applied to each trade to cover commissions and slippage. Table 5-2 summarizes the ty of sjstem, calculation method, and trading rules.

Varjing the Speed of the Trend

Each sjstem must be tested with a wide range of fast and slow trends. For sjstems 1 through 4, the period, or number of dajs, n, determines the trend speed. For the swing

Summary oT Six MorTV-endlng Syscenns

- closing price:

MA - @Surn(c1o5e MA - @Averagecc MAchange - MA -

up or down)

2. ExpoitenElal amoothing

EXP is a weistnins <)1 *e recent closing pri» by a p r tag-.s (called che o th... ii„.,.,inO

<Ehe crsnd [urns up er downj

Bu). when The is up (slope is

poiiljvs). and spW when the = ic rfown (sic.p« is negative)

©lowest<low[i].r,-i

higher than the pi

X. Swinsbmicoi

Siena occur high low previ swins hih or low pHce.

begins with reversal of a

mi entage value.

Buy when Ihe current hl exceeds ihe previous swinj high Sell when the current low flls below the previous swing low

when the c< high or low price exceeds the previous high or low. which is formed by filling boxes of a pres

high :

(cepi thai new highs and lows e recorded only If th exceed a

-eiat box siie. A reversal ii Buy when the current high

>corded if a minimum of 3 boxes exceeds the previous PF swi in be filled, not including ihe high by at least one box. Sell

the pi

.s PF swing low by a,

nil 111

111 I

flfl llll

II, III

= " =

Si s: =; - - = -

breakout, a percentage swing makes it more or less sensitive to price movements. Pointand-figure uses a constant box size, which is then multiplied by 3 to get the reversal size

Tests shown in Table 5-3 for the Eurodollar confirm a broad range of trading from very short-term to very long-term. The number of trades varies from over 500 in 10 years (about 2,500 dajs) to 6 trades in 10 years. Some showed a fairly high minimum number of trades for the longest periods.

Different Trend Philosophies

The movmg average, exponential, and linear regression are time-series processes; that is. eadi dsy is a small part of the total calculation. This can make it very difficult for a single large price jump to cause a change of signal. For very fast trends, eadi price will have more importance than for slower trends.

The N-dsy breakout, swing breakout, and point-and-figure sjstems all depend on a price breakout. A signal occurs when the maricet makes a new high or low. Therefore, a single big move often causes a new signal. Breakout sjstems can be considered event driven, wliile a moving average sjstem requires a trend to evolve.

Different Ride Profiles

Each of the six sjstems fall into particular categories of risk profiles Moving Averages and Exponential Smoothing

Both the moving average and exponential smoothing retum a weighted average. The simple moving average uses equal weighting, and the exponential gives more importance to the most recent value, according to the smoothing constant, s.

Both sjstems are considered conservation-of-capital methods because they keep losses small. When current price levels are trading near the n-dsy average, it takes very litUe to change the trend from up to down and back to up. Although this is annojing to traders, it is part of the conservationof-capital benefit. Table 5-3 diows that the minimum number of trades were 42 to 44 for these sjstems, far greater than the other four strategies. The whipsaws seem to increase as the trends become very slow, which is the reason for using a filter, or band, in most trend sjstems. In Table 5-3 the number of trades doubled when the trend stowed fran 150 to 200 dsjs.

Breakout Systems

The swmg breakout, point-and-figure, and N-dsy breakout sjstems can all be grouped as high-ride strategies. Once a buy signal has occurred because of new high prices, the trade is reversed only when new low prices occur. The initial risk is

breakout ride = highest high - lowest low

for the N-dsy period. The rate at which ride declines is different for eadi of the sjdems. The swing breakout and point and figure sjstems reduce ride eadi time a new swing low occurs only if that low is higher than the previous swing low (see Figure 5-5). The N-dsy breakout reduces ride if the oldest dsjs, which are dropped off, were also the lowest prices. Otherwise, its risk is the same as the other breakout sjstems.

Linear Regression Risk

The linear regression method, which creates a sU-aight-line forecast, is very different from other trending sjstems. It does not change direction when prices move up or down through the trendline (as in the smoothing sjdems). It is possible that a very large, but fast price drop, which recovers quickly, will not cause a long position signal to reverse.

FIGURE 5-5 Swing breakout risk pattem. ,\s prices move lower, the swing highs become the point at which signals ajs the same for point-and-figure.