An ADX Burst Trend-Seeking System

We have assumed that the market was about to trend in both the 65sma-3cc and the CB-PB systems, although we did not actually verify that the market was trending because it is difficult to measure trendiness consis­tently. As was shown in the discussion in chapter 3 on the range action verification index, market momentum is often a good measure of trend­iness. Unfortunately, a certain amount of smoothing is essential to mini­mize noise in the indicator, and this smoothing usually causes undesir­able lags in indicator response.

Figure 4.23 shows the March 1993 U.S. T-bond contract trending upward nicely from December 1992 through March 1993. The indica­tor under the daily bars is the 18-day average directional index. ADX measures the amount of activity outside the previous bar over a given period; a strong trend usually leads to a rising ADX line. An ADX read­ing above 20 is considered to indicate a trend, but the ADX is a lagging indicator, and there is little significance to any particular indicator value.

ADX is closely related to double-smoothed absolute momentum, and hence will often have quirky lags. The ADX will often seem to be late in signaling a trend, and choppy markets will not follow through in the original direction that caused the ADX to rise. In fact, the market can reverse strongly, and the ADX will keep on rising.

During a strong trend, as markets make big daily moves in the di­rection of the trend, the daily ADX momentum can "pop" over 1.0


112 Developing New Trading Systems

112*16

,.t

106*08

-103*04

Dec 93 Feb Mar

Figure 4.23 A rising 18-day ADX can be a good indicator of a trending market.

point, an ADX "burst." Figure 4.24 shows the March 1993 U.S. bond contract with the histogram of the ADX burst superimposed on the 18-day ADX line. As the trend accelerates, the daily ADX changes are more than 1, and you can see relatively large bars associated with this ADX

112*16

/

109*12

-106*08

-103*04

Figure 4.24 The histogram of ADX burst momentum shows daily changes greater than 1.


An ADX Burst Trend-Seeking System113

burst activity. Now you can build a trading system using this idea as shown in Figure 4.25, where the entries are circled.

Obviously, the ADX burst indicates accelerating momentum. So, here the design philosophy has changed to begin with a check that in­creases the odds of success of a trend-following strategy. Notice that the ADX burst is itself triggering the trade, and that the ADX is not acting as a filter. For reference, you can look up a similar system in Lucas and Le Beau (see bibliography for reference). Our goal is to take the trade in the direction of the short-term trend. If the 3-day SMA is greater than the 12-day SMA, then the trend is up, and vice versa. Table 4.12 shows the results using a simple 20-day exit strategy and allowing $100 for slip­page and commissions, over all available data from Janury 1, 1975, through July 10, 1995.

The rather large profit factor suggests that the entries are effective in identifying profitable trades, so that an ADX burst is a good entry into strong trends. The profit factor is overestimated here to some de­gree because we are using continuous contract data. The results can be improved with multiple contracts, and you can try a variety of other exit strategies.

If you compare the number of trades here to that for the 65sma-3cc system, you will find that you have fewer entries, suggesting that the ADX burst is working as both a trade filter and a trigger. For example,


112*16
•I09--I2
-106*08
.103*04

Figure 4.25 A trading system triggered by ADX burst with daily momentum changes more than 1.


114 Developing New Trading Systems Table 4.12 ADX burst system performance with $5,000 initial stop

Market Profit ($) Winning Trades; Total Trades; Winning Percentage Profit Factor (Gross Profit; Cross Loss) Average Trade (S) Maximum Intraday Drawdown ($)
British pound 40,531 34;75;45 1.39 -25,113
Canadian dollar 6,830 20;56;36 1.28 -7,060
Coffee 137,014 29;75;39 3.86 -21,225
Copper, high- 4,770 15;25;60 1.29 -5,970
grade                    
Corn 22,269 37;70;53 2.58 -3,356
Cotton 72,770 32;64;50 3.49 1,138 -4,860
Crude oil 10,590 21;54;38 1.44 -13,400
Deutsche mark 63,300 40;73;55 3.49 -8,675
Gold 4,770 31;84;37 1.04 -27,450
Heating oil 52,469 30;56;54 2.61 -7,850
Japanese yen 63,450 37;69;54 2.35 -18,050
Live hogs 20,080 36;75;48 1.65 -6,140
Orange juice 25,013 29;80;36 1.63 -12,692
Soybeans 38,606 31;73;42 1.81 -10,713
S&P-500 -28,650 20;55;36 0.79 -520 -65,815
Swiss franc 76,238 35;68;51 2.75 1,121 -8,075
U.S. bond 54,531 27;60;45 2.56 -11,306

 


Average

39,093

2.12

 


 


this system was in the market about 35 to 45 percent of the time, indi­cating it has a rather large "neutral zone." A trading system with a neu­tral zone is out of the market unless it rises above stiff entry barriers. The 65sma-3cc system is always in the market, and is a reversal-type sys­tem, whereas the ADX burst system steps aside 55 to 65 percent of the time.

We used a wide initial stop of $5,000 in these calculations to isolate the performance of the system. Table 4.13 includes performance data on selected markets with an initial stop of $1,500. The performance with the two different initial stops was generally similar.

One of the quirks of the ADX burst system is that it will often get in late, near the tops or bottoms of short but swift moves (see Figure 4.26). Such moves fire its entry signals, but the capricious market fails to follow through with a trend in the advertised direction. Hence, you


An ADX Burst Trend-Seeking System115 Table 4.13 ADX burst system performance with a $1,500 initial stop

                    Maximum    
Market Profit (S) Number of Trades Percen­tage of Wins Average Trade (S) Intraday Drawdown (S) Profit Factor
British pound Coffee 64,438 148,584 82 85 39 33 744 1,749 -18,719 -13,851 1.82 3.07
Cotton 66,800 1,012 -6,015 2.95
Crude oil 6,070 -13,440 1.24
Deutsche mark 62,088 -8,457 3.34
S&P-500 19,160 -33,675 1.24
Swiss franc 61,575 -9,125 2.88
U.S. bond 40,556 -12,944 1.74

 

should always trade a system such as this one with a preplaced stop loss order.

In summary, the ADX burst system provides entries into strong trends. It tests well across many markets and over long time periods. The system has a large neutral zone, so it is in the market only 3 5 to 45 percent of the time. It differs from the 65sma-3cc system, which is

    •94*12
HI]         •93*24
        i, ,, •93*04
'mt i     I'll,,!1'!.11 -92*16 •91*28
            0 1    
                •91*08
        \ "it     .90*20 -90*00
i | . -i        
        -89*12
^rii1        
^ -88*24
    -88*04
               

 


Apt

May

 


Figure 4.26 The June 1990 U.S. bond contract sells off beyond a trading range to make a new low with good momentum. The system kicks in with a short. The bond market soon reverses, to get back into the prior consolidation region.


116 Developing New Trading Systems

always in the market, and does not have a trend filter. You can use it to enter trades or increase the position in those markets. You can derive other variations using different values of the ADX burst, the look-back period for the burst calculations, and other exit strategies.