Channel Breakout-Pull Back Pattern
This section discusses a trading system based on a pattern observed in mature markets, that is, markets with a large volume of institutional activity. In these markets, the big players have a tendency to fade market moves. Thus, they will resist advances and support declines. For example, when a market makes a new 20-day high, many big players will short it heavily, and push the market back into the previous consolidation. If the fundamental forces underlying the market are strong, the up trend will resume after a brief consolidation. A trading system that trades the long side only, by going long during the pull back after new 20-day highs, is called the channel breakout-pull back (CB-PB) system.
We begin with a few examples of how the CB-PB system works, and show the actual code used for the tests. Next, we test the basic CB-PB entry strategy across 22 markets to illustrate the general validity of the idea. Then, we discuss three different exit strategies to show how you can convert the same entry strategy into vastly different trading systems. These systems vary from a short-term system, which is in the market for 7 to 9 days, to a long-term trend-following system. We will also explore the effect of using a $1,500 "close" initial stop versus a $5,000 "wide" stop. The analysis focuses on the following mature markets: coffee, Eurodollar, Japanese yen, Swiss franc, S&P-500, 10-year T-Note, and the U.S. bond.
The channel breakout-pull back pattern is for long trades only. The assumptions underlying this system are:
1. The market will begin an uptrend after the consolidation ends, because it has recently made a new 20-day high.
2. The entry during the consolidation is a low-risk entry point.
3. Exits could be placed at the nearby 20-day high, by using trailing stops, or by exiting after .r-days in the trade.
The reality is that markets may have an extended consolidation after making a new 20-day high, or could even make new 20-day lows.
Developing New Trading Systems
Hence, a bias to the long side may be correct only 50 to 70 percent of the rime. It is also difficult to find consistent exits, since the markets do not follow the same script every rime. Hence, another difficulty with the CB-PB system is finding a consistent exit strategy. A third area of difficulty is where to place the initial stop. If the market rolls over and starts a new downtrend, then an initial stop is critical for risk management and loss control, whether it is a simple-dollar stop or a volatility-based stop.
The first example of the CB-PB pattern uses the March 1995 deut-sche mark contract. Figure 4.19 shows the daily bars and, superimposed on the bars, the 20-bar trading range. The 20-day range lines have a 13-rick barrier added to both the lines to filter out some false breakouts. The chart shows that the deutsche mark broke above its 20-day range in December 1995 and then consolidated for 7 days before moving higher. Upon moving higher, it made a higher high, and consolidated again.
Ideally, we would like to buy some rime during the pullback, but we do not know how long the pullback will last. Hence, the problem is how to specify that a pullback has occurred. During the pullback, markets often also make new 5-day lows. Hence, we can define this breakout and pullback long entry rule as follows: the market must make a new 20-day high, and then define a 5-day low in the next 7 days. Once it forms a
-66.50
-66.00
| 65.50 -65.00 |
-64,50
-64,00
-63.50
-63.00
Figure 4.19 The deutsche mark pulls back after making a new 20-day high. The goal is to buy after the pullback. The 20-day price channel is shown for visual reference.
Channel Breakout-Pull Back Pattern103
5-day low, buy on the open the next trading day. These choices are arbitrary, and you can experiment with these numbers. For example, we can buy on the close instead of on the open after the market forms a 5-day low following a 20-day high.
We now need an exit condition to evaluate this entry rule. To keep it simple, we will exit on the close of the w-th day in the trade, with n=5 for short-term systems and n=50 for intermediate systems. Again, these numerical values are arbitrary. You may try other values, such as a 3-day low instead of a 5-day low.
Using the Omega Research TradeStation Power Editor™, the rule appears, in part, as:
Input: Xdays (14);
If Highest Bar(High,20)[1] < 7 and Low < Lowest
(Low,5)[l] then buy tomorrow on the open:
If BarsSinceEntry - Xdays then exitlong at the close:
The first line defines "Xdays" as an input-variable with a default value of 14 days. You can change this value during testing. The Highest Bar function returns the number of bars (trading days) since the 20-day highest high. The second line first checks if 7 or fewer days have elapsed since the new 20-day high. Then, it checks if today's low is lower than the previous 5-day low (i.e., a new 5-day low). If both conditions are true, then you can buy tomorrow on the open. By default, this system will buy one contract. The third line is the exit condition, which says that if today is the r-th day since entry, then exit the long trade at the market on the close. This system will fill the long trade at the opening price of the entry day, and at the closing price of the exit day.
There is a quirk in how the Highest Bar function works. The function counts 20 days back from the day it is testing. Hence, the function will occasionally give a signal that does not work off the highest high as intended. Hence, to accurately pick off the highest high of the last 20 days, the rule should say Highest Bar(High,27)[l]. However, the difference in the results over the long run is insignificant.
Figure 4.20 shows that the March 1995 deutsche mark chart with a 14-day exit worked well. The first breakout occurred on December 28, 1994, and the pullback entry occurred on January 9, 1995, at the open of 64.11, which was the exact low of the ensuing 14 days. The exit was on the close of January 30, 1995, at 66.52, for a profit of $2,913, after allowing $100 for slippage and commissions. The next entry occurred on
104 Developing New Trading Systems
Figure 4.20 The CP-PB strategy gave good trades with low-risk entry points.
February 1, 1995, on the open at 65.65. The low of the trade occurred four days later at 65.07, for a 58-dck risk of $725. The exit was on the close of February 23, 1995, at 68.19. The nominal profit was $3,075.
Thus, the CB-PB system generated low-risk entries into an emerging up trend in the March 1995 deutsche mark contract. The exit on the 14th day was a lucky choice for this chart. You could use a number based on your individual preference just as well.
Note here that we specified a generic entry pattern with no specific assumptions about DM price patterns. The exit was again arbitrary. Of course, if you had exited on the 5th-day close instead of the 14th-day close, the profits would have been smaller. Note that the CB-PB pattern offers a relatively low-risk entry method. You can use it as a short-term system or a long-term system by simply varying the exit strategy.
So far, the exit strategy has been trend-following in nature, with some variation based on the actual day of the exit. For example, we could vary the exit from 5 days to 50 days and get completely different results. However, we will never make the "perfect" choice of x days. We can anticipate market action in a different way that does not use time as the exit signal. Instead, we will use a price we already know. Since we are buying a pullback, it is plausible to assume that the market will retest the recent 20-day high. Hence, we can write an exit signal that buys the
Channel Breakout-Pull Back Pattern105
pullback and exits the retest of the recent high. Here is how we would write the new system variation in TradeStation™:
If Highest Bar(High.20)[l] < 7 and Low < Lowest
(Low.5)1:1] then buy tomorrow on the open;
Exitlong at highest(h,20)[1] limit;
The first line of the CB-PB rule is exactly the same as before. The second line specifies a long exit for tomorrow with a limit order at the most recent 20-day high. This turned out to be the "perfect" model for the December 1995 S&P-500 contract. There were 12 winning trades in a row, with a total profit of $50,000 (see Figure 4.21).
The noteworthy feature here is that we started with the DM contract, using very general price patterns, and arrived at an intriguing short-term system, which performs particularly well in choppy uptrends. We made no contract-specific assumptions, and captured a general market behavior that we can expect to see in every market in the future. The CB-PB entry with an exit at a recent high works well in consolidations.
A M J J A S 0 N
Figure 4.21 The CP-PB model with exit at the recent 20-day high using limit orders produced 12 winning trades in a row for a nominal profit of $50,000.
106 Developing New Trading Systems
Another exit strategy involves a trailing stop, but one that will not cut off long trends prematurely. Hence, we will exit at the lowest low of the last 40 days. This will convert CB-PB into a long-term trend-following system.
If Highest Bar(H1gh,20)[l] < 7 and Low < Lowest
(Low,5)[l] then buy tomorrow on the open;
exitlong at lowestClow. 40)[1] - 1 point stop;
The CB-PB entry rule remains intact. The second line exits on a stop set one tick below the trailing 40-bar (trading days) low. You can see that this will become a trend-following exit. Our initial stop will close out our trade should the market head lower. The trailing stop at the 40-bar low will keep us in the trade through minor consolidations.
Notice how we took an intuitive understanding of a market pattern and adapted it to three different exit philosophies to meet specific trading preferences. Remember you could use it as a short term system by exiting at the recent high. You could exit on the close of the n-th day in the trade, for short- or intermediate-term trading. Or you could use a trailing stop. Each exit produces a trading system with different characteristics off the same entry signal. These are the types of modifications you should consider as you look at trading systems. Figure 4.22 from the March 1995 U.S. bond market will help you visualize the three exit strategies.
Now let us take a closer look at the entry signal, to see if it is any better than a random entry system. Following the suggestion of Le Beau and Lucas (see bibliography), we will try to isolate the effect of this CB-PB entry signal.
We test the CB-PB entry signal with exit on the close of the n-th day (n=5, 10, 15, and 20), without stops and assuming no slippage or commission costs. Le Beau and Lucas suggest that if the entry signal is performing better than a random system, it should result in at least 55 percent profitable trades over a range of markets. They tested only 6 years of data and 6 markets to measure a signal's ability to perform better than random. Here we use 22 markets and continuous contracts using all available data from January 1, 1975, through July 10, 1995. This should be a severe test of this entry signal, and our goal is to check if it is consistently profitable more than 55 percent of the time.
Table 4.7, page 108, shows that about 55 percent of all CB-PB entries were profitable. Hence, you can be reasonably confident that the CB-PB entry signal provides better than random entries. You can now
|
Channel Breakout-Pull Back Pattern107
-106*08
-104*22
-103*04
-101*18
-100*00
-98*14
-96*28
Figure 4.22 The CB-PB gave a low risk entry into the new trend for the March 1995 U.S. bond contract.
marry this entry signal to a variety of risk control and exit strategies to fashion a trading system that fits your trading mentality.
The first exit strategy is simply to exit on the close of the w-th day in the trade. You are making the working assumption that the market is going to trend after the entry signal. Hence, consider now the CB-PB entry using continuous contracts, $1,500 initial stop, and allowing $100 for commissions and slippage. As discussed at the beginning of this section, we are focusing on "mature" markets. Let us consider the case when we exit the long trade on the close of the fifth day. The test uses all available data from January 1, 1975, through July 10, 1995.
The results of exiting on the fifth day of the trade are not impressive (see Table 4.8, page 109). Since we are buying the markets during a consolidation, most of them have not done much in the 5 days after entry. Hence, we should consider holding on to the long trade for a little while longer.
Consider what happens if we hold the long position for 50 days, exiting on the close. The conditions for the test are identical to those for Table 4.8. Table 4.9, page 109, shows there is a dramatic improvement in performance with n = 50 days. The average profit per market has increased three-fold, and the profit factor is up 46 percent. Thus, our basic assumption that the market will trend after the consolidation seems to work well about 39 percent of the time on these markets. Thus, we have
108 Developing New Trading Systems
Table 4.7 Percent winning trades for CB-PB entry signal calculated over all available data from January 1, 1975, through July 10, 1995
| Market | 5-Day Exit | 10-Day Exit | 15-Day Exit | 20-Day Exit |
| British pound | ||||
| Canadian dollar | ||||
| Coffee | ||||
| Copper | ||||
| Corn | ||||
| Cotton | ||||
| Crude oil | ||||
| Deutsche mark | ||||
| Eurodollar | ||||
| Gold | ||||
| Heating oil | ||||
| Japanese yen | ||||
| Live hogs | ||||
| Orange juice | ||||
| Silver | ||||
| Soybeans | ||||
| S&P-500 | ||||
| Sugar | ||||
| Swiss franc | ||||
| 10-year T-note | ||||
| U.S. bond | ||||
| Wheat | ||||
| Average |
converted our anemic short-term system into an interesting intermediate term system by exiting on the close of the fiftieth day.
We have previously stated that the initial stop should depend on market volatility. For example, the $1,500 stop may be "too close" given the volatility of the S&P-500 market. For the CB-PB system with exit on the 50th day using a $5,000 initial stop instead of the $1,500 initial stop, the profits dropped for all markets in Table 4.9 except S&P-500. Profits for S&P-500 increased to $141,840 on just 55 trades with 56 percent winners, a $2,579 average trade. The maximum drawdown was -$24,795, with the profit factor increasing to 2.29 from 1.62. Hence, the initial stop will influence overall system performance.
We can continue to explore the long-term nature of this entry by using a trailing stop. We know from Table 4.9 that we should use a trailing stop that will allow trends to develop. Hence, let us arbitrarily spec-
Channel Breakout-Pull Back Pattern109
Table 4.8 CB-PB long trades with exit on the 5th day using $1,500 initial stop, tested on all available data from January 1, 1975, through July 10, 1995
| Maximum | ||||||
| Market | Profit ($) | Number of Trades | Percentage of Wins | Average Trade ($) | Intraday Drawdown (S) | Profit Factor |
| Eurodollar | 6,050 | -4,350 | 1.27 | |||
| Japanese yen | 27,450 | -9,863 | 1.63 | |||
| Coffee | -11,273 | -94 | -23,500 | 0.86 | ||
| S&P-SOO | 69,330 | -19,640 | 1.42 | |||
| Swiss franc | -4,988 | -42 | -1 7,913 | 0.94 | ||
| 10-year T-note | 18,831 | -8,756 | 1.39 | |||
| U.S. bond | 27,306 | -13,219 | 1.45 | |||
| Average | 18,958 | -13,892 | 1.28 |
ify an exit on the lowest low of the last 40 days; this should convert the intermediate system into a long-term trading system. As before, we will use $1,500 initial stop and allow $100 slippage and commissions.
Table 4.10 shows the long-term performance of this entry with a profit factor of nearly 3 and an average trade of $1,082. The ratio of net profits to drawdown is more than 4.5. These numbers suggest that you
Table 4.9 CB-PB long trades with exit on the fiftieth day, using $1,500 initial stop, tested on all available data from January 1, 1975, through July 10, 1995
| Maximum | ||||||
| Market | Profit ($) | Number of Trades | Percentage of Wins | Average Trade (S) | Intraday Drawdown (S) | Profit Factor |
| Eurodollar | 21,875 | -8525 | 1.74 | |||
| Japanese yen | 76,613 | 1,473 | -11,525 | 2.69 | ||
| Coffee | 27,434 | -18,719 | 1.33 | |||
| S&P-500 | 86,085 | -26,475 | 1.62 | |||
| Swiss franc | 52,889 | -13,900 | 1.81 | |||
| 10-year T-note | 49,799 | -9,575 | 1.98 | |||
| U.S. bond | 63,094 | -14,169 | 1.95 |
| Average |
| 53,970 |
| -14,698 |
| 1.87 |
110 Developing New Trading Systems
Table 4.10 CB-PB long trades with exit on a trailing stop at the 40-day low, using $1,500 initial stop, tested on all available data from January 1, 1975, through )uly 10, 1995
| Maximum | ||||||
| Market | Profit ($) | Number of Trades | Percentage of Wins | Average Trade ($) | Intraday Drawdown ($) | Profit Factor |
| Eurodollar | 32,200 | -3,375 | 3.65 | |||
| Japanese yen | 70,419 | 2,063 | -7,112 | 4.39 | ||
| coffee | 53,928 | -24,020 | 2.00 | |||
| SScP-500 | 85,200 | -25,480 | 1.41 | |||
| Swiss franc | 55,200 | -11,550 | 2.42 | |||
| 10-year T-note | 57,250 | 1,123 | -8,038 | 3.39 | ||
| U.S. bond | 62,513 | 1,158 | -11,475 | 2.13 |
| Average |
| 59,530 |
| 1,082 |
| -13,007 |
| 2.77 |
can take the same entry and make it into a strong long-term trend-following system.
Let us now take the CB-PB entry and attach it to an exit at the recent 20-day high. It is reasonable to assume that the market will retest the recent 20-day highs as part of the backing and filling during the consolidation. Table 4.11 summarizes the test results using a $1,500 initial stop and a $100 allowance for slippage and commissions.
Table 4.11 CB-PB long trades with exit at the recent 20-day highs^on a limit, using $1,500 initial stop, tested on all available data from January 1, 1975, through July 10, 1995
| Maximum | |||||||
| Market | Profit (S) | Number of Trades | Percentage of Wins | Average Trade (S) | Intraday Drawdown ($) | Number of Days in Wins | Profit Factor |
| Eurodollar | 7,250 | -8,750 | 1.24 | ||||
| Japanese | 17,200 | -11,225 | 1.30 | ||||
| yen | |||||||
| Coffee | -7,751 | -66 | -24,463 | 0.93 | |||
| S&P-500 | 48,860 | -25,070 | 1.25 | ||||
| Swiss franc | -5,963 | -51 | -16,625 | 0.97 | |||
| 10-year T- | 26,781 | -8,388 | 1.42 | ||||
| note | |||||||
| U.S. bond | 37,306 | -10,856 | 1.47 | ||||
| Average | 17,669 | -17,377 | 1.22 |
An ADX Burst Trend-Seeking System111
The CB-PB system with an exit at the recent 20-day high was interesting only on the Eurodollar, S&P-500, 10-year T-note, and U.S. bond markets. The large proportion of winning trades makes this exit particularly attractive. Notice that the length of the average winning trade was only 9 days.
You can develop other variations of this strategy. For example, one of the design features of the CB-PB system is that we want a low risk entry point into long trades. Hence, you can use a multicontract trading strategy to improve performance. Another approach would be to add a filter to reduce the number of trades.
Thus, the CB-PB system has a flexible entry to suit many trading styles. The CB-PB strategy is more profitable with an intermediate to long-term trading strategy. A short-term approach worked on a few active markets. Note also how we can develop different systems from the same entry signal by changing the exit strategy.