Tradingpatterns.com Michael Harris article

Trend Trading with Short-Term Patterns

By Michael Harris

Have you ever "doubled up" only to see the stock price continue to decline?

The rewards of trend trading can far exceed those of constantly fighting for an extra penny on a scalp trade, but the approach has inherent difficulties. By breaking down longer-term trends into shorter-term patterns, traders can enjoy larger-scale profit potential without giving up shorter-term trading advantages.

Some traders are often so immersed in their frenetic routines on the intraday time frame they fail to notice the potential profits available in short-term or intermediate trends. These price moves can be significant and their profits far greater than those obtained by chasing one-minute price fluctuations.

However, trend-following has fallen increasingly out of favor in recent years. There are a few reasons traders are discouraged from attempting to trade trends. First, trends can be defined only in hindsight and there are no robust trend-trading indicators. Also, the psychological burden of trend trading – which can require waiting out long holding times and large drawdowns – is high.

Combining shorter-term price patterns, however, can sometimes make it possible to capture trends without undue exposure to these drawbacks.

A price trend is probably one of the most elusive concepts in trading because a trend can only be identified after a significant portion of it has already formed. More importantly, every price level in a trend is a potential reversal point, making it impossible to know if prices will trend up or down at any given time.

Also, any indication of a longer-term directional bias cannot be determined from analyzing price data alone. This is because buyers and sellers match exactly at every price point; the only thing that differs is "price concession." When buyers concede to higher offers, prices move up; the opposite occurs when sellers concede to lower bids. However, there is no way of knowing traders' longer-term motives. Among other factors, short-selling and bluffing blur the market picture, resulting in "noisy" market conditions. Traders who understand how noise affects technical indicators, especially trend-following ones, know these tools have little value as the basis of trend-trading systems with a chance to be consistently profitable over time.

But in addition to the theoretical problems that explain why commonly used technical trend-following methods tend to fail, there are also some disturbing realities every trader becomes aware of at some point.

Sudden trend reversals can cause devastating reductions of open position profit, or even turn a profitable trade into a loser. Understandably, these events can adversely impact a trader's psyche, especially a trader with limited experience. Even veteran traders are not immune to the stress of holding open positions for extended periods of time and being subject to high volatility and adverse gap openings.

At the other end of the spectrum, many short-term traders become addicted to pulling the trigger and pocketing small profits after minimal favorable price moves. When it comes to trend trading, though, patience is a virtue.

It is possible to participate in trends by exploiting certain features of short-term price patterns. Doing so reduces both the uncertainties of traditional trend-following techniques as well as the psychological burden they impose.

The concept: short-term trading systems consisting of sub-systems based on price patterns can exhibit trend-following properties. It is crucial to understand that such systems do not try to predict trend direction or duration. Consequently, the trend-following capacity of such systems is an added benefit to the price pattern's ability to follow market swings.

When using this method, the traditional goal of trend-following – finding a tool or technique that will identify long-term price trends – is transformed into a search for a combination of short-term price patterns that generate signals with the ability to capture a good portion of a price trend.

The trade signals based on short-term patterns found within a longer-term trend produced much better results than those triggered by MACD crossovers (buying and selling when the solid indicator line crosses above and below, respectively, the dotted signal line).

Source: MetaStock

Figure 1 shows a short-term trend that developed in the S&P 500 index-tracking stock (SPY). The trend started in the middle of September 2002 and lasted until the end of that year.

The total profit realized by the four hypothetical signals, shown for illustration purposes only, would have produced a much greater total gain than any profit realized by using a simple indicator, such as an MACD (superimposed on price).

Signals 1 through 4 are "successive," which means they do not overlap – i.e., each pattern signal exits with a profit or loss before the next signal is generated. The properties of successive signal patterns makes it possible to create effective trading methods with flexible risk control schemes.

In describing this trading method, we will illustrate the underlying concept and basic technique for following trends using short-term price patterns, rather than focusing on patterns themselves. The goal is to find patterns with certain length, frequency and profitability characteristics; such patterns can take many specific forms.

The price patterns in question are distinct price bar formations that generate signals based on specific market conditions. Although the signals of different patterns may sometimes overlap, resulting in "coincident" or "clustered" pattern formations, there is always a good chance that during an intermediate or long-term trend several "successive" (non-overlapping) signals will form. This is the main advantage of basing a system on several different price patterns rather than a single trading indicator. Different patterns generate different signals because of their unique price bar formations, thus increasing the chance of capturing a trend. An example of a price pattern found in this study is shown in Figure 2.

Price patterns can be clearly defined by the relationships between price bars and their open, high, low and closing prices. The series of relationships shown on the left describes the pattern on the right.

Another reason to use several price patterns is diversification of risk. Each pattern can be treated as a unique trading indicator. For this experiment each pattern was required to have sufficiently high historical profitability, a sufficiently high number of trades and a low number of maximum consecutive losses.

With short-term price patterns, the historical profitability depends on the ratio of the profit target to the stop-loss (e.g. 4 percent vs. 2 percent, which is a 2:1 ratio), which in turn determines the average-win to average-loss ratio. For instance, if the ratio of the profit target to the stop-loss is 1:1, the strategy must have at least a 50 percent "profitability" (the ratio of the number of winners to the total number of trades) to break even. As the demand for higher profitability increases, the chances of finding viable patterns diminishes. Similarly as the number of historical trades increases, the probability of finding patterns decreases.

The number of maximum consecutive losers directly impacts a price pattern's maximum drawdown and, consequently, a trading system that uses it. The theoretical minimum is zero, corresponding to 100-percent profitability (no consecutive losers). A maximum of three to five consecutive losing trades is realistic for price patterns. However, you must experiment with such performance settings before designing a trading system.

The following methodology and pattern criteria are suggested starting points for individual research and testing. The goal is to find patterns that are viable individually, but which also work together to capture trend moves if they occur. Such price patterns can be identified either visually on price charts, or preferably through software that automates the process based on user defined criteria.

Depending on your specific trading goals and risk level, your parameters may fall outside these boundaries.

1.	A typical pattern should generate at least 26 to 30 trades over the testing period to support the statistical significance of back-testing results. No one should expect a pattern that has formed only a couple of times in the past to represent a good trading opportunity in the future.

2.	Although patterns may overlap at times, they should generate a sufficient number of successive (non-overlapping) trade signals. A way to check this is to test the patterns individually, record the number of trades signaled by each and then calculate the total for all patterns; let's call that number "Ti." Then test the composite trading system (with all patterns working in sync), and record the number of trades; let's call this number "Ts."

	The ratio Ts/Ti must be as close to 1.00 as possible (by definition it cannot be above this number). A ratio of 0.5 would indicate that at least 50 percent of the signals do not overlap, which is a good starting point. A ratio of 0.7 is a better objective; from experience, it produces good results.

3.	The typical pattern length is two to 10 bars. However, using patterns that consist of more than six bars usually reduces the number of historical trades and thus the significance of a pattern's performance (see #1).

4.	The typical trade length is from one bar to several bars, depending on market volatility and the specific profit target and stop-loss objectives used.

5.	A pattern's reward/risk ratio and winning percentage are interdependent. The higher the reward/risk ratio (related to the average winning trade vs. the average losing trade), the lower the winning percentage required to achieve profitability. Conversely, the higher the winning percentage, the lower the reward/risk ratio required for profitability.

Searching for patterns with very high reward/risk ratios and/or winning percentages will make it difficult to find patterns that appear frequently enough to produce statistically reliable results in testing.

In the following examples, the reward/risk ratio criteria is 1:1 – that is, each pattern's profit target and stop-loss level are the same size (which means the winning percentage must be above 50 percent to produce a profit). In this case, we will use a minimum winning percentage of 66 percent, to compensate for slippage and commissions and still leave room for a profit.

Figure 3 shows the result for a search (using the APS software; see note at end of article) spanning January 1990 to Feb. 14, 2003, for price patterns with the following performance characteristics:

MSFT
> 66%
> 30
< 4
open of next day
= 7% of entry price

Short-term price patterns can be combined to create successive signals that capture trend moves. Pattern-finding software (in this case, the APS program) discovered short-term patters that met certain performance criteria. These patterns were then back-tested, producting the results in Table 1.

Legend: Trade on designates whether the trade entry executed on the open or close. • PL is the percent profitability of patterns suitable for long positions. In this case PS = 100 - PL. • PS is the percent profitability of patterns suitable for short positions. In this case PL = 100 - PS. • Trades is the number of trades. • CL is the number of maximum consecutive losers. • Type is either Long or short. • Target shows the profit target value used in the search. • Stop shows the stop-loss value used in the search. • C indicates the type of profit target and stop-loss; "%" stands for percentages of entry price.

Source: Automatic Pattern Search (APS)

Testing a combination of five simple patterns (such as the one shown in Figure 2) produced the following results in Microsoft over a 13-year period.

Source: MetaStock

Five patterns (including the one shown in Figure 2) met the specified criteria. Table 1 shows the back-testing results for the entire testing period for a trading system using these five price patterns as its sub-system.

Figure 4 shows trades that occurred during a short-term trend from July 2002 to August 2002. During that period, the system generated three profitable signals and captured a good portion of the overall move.

Trade entry	Trade exit	Profit/Loss
7/24/2002	7/25/2002	$3.70
8/5/2002	8/8/2002	$3.10
8/16/2002	8/22/2002	$3.46

The patterns that produced the performance results in Table 1 triggered three winning trades in the one-month period shown here.

Source: MetaStock

The net profit for those three trades was $10.26, while the net buy-and-hold gain for the entire period was about $12. Overall, the three trades captured approximately 86 percent of the trend.

Historically, systems based on technical indicators have a difficult time rivaling the performance of the kind of price patterns used in this example. For example, a simple 5-30 moving average crossover (buying when the five-bar moving average crossed above the 30-bar moving average, and selling on the opposite condition) would result in a losing trade, as shown in Figure 4. A system using the moving average convergence-divergence (MACD) indicator, with a four-day exponential moving average of the MACD as the trigger line, generated two trades – one losing, one winning – as shown in Figure 5. The trades were:

Trade entry	Trade exit	Profit/Loss
7/31/2002	8/5/2002	-$2.44
8/8/2002	8/29/2002	$3.89

In contrast to Figure 4, the two signals that would have been triggered by the MACD (overlaid on price) would have resulted in one winning trade and one losing trade.

Source: MetaStock

The exit signal of the second trade in Figure 5 underscores a problem with the MACD system: The signal reduces profit substantially because the indicator doesn't react swiftly enough to the price correction that reverses the uptrend. This is a common problem when using any "smoothing" (i.e., moving average-based) indicator to signal trades.

Traders have used trailing stops to improve the performance of indicator-based signals, but these stops are adversely affected by trend volatility. For example, if volatility is high, an exit signal might be generated during the main portion of the trend.

The ability of this sample system – which has been created by combining five price patterns – to follow short-term trends in MSFT is further demonstrated in Figure 6. In this case, the trend began in June 1999 and peaked in mid-July. The trading system generated three successive profitable buy signals:

Trade entry	Trade exit	Profit/Loss
6/1/1999	6/21/1999	$5.65
6/29/1999	7/6/1999	$6.06
7/12/1999	7/16/1999	$6.52

The three trades from this period captured approximately 73 percent of the trend move that occured.

Source: MetaStock

Net profit for the three trades was $18.23 vs. a total potential profit for the trend of $25. This means the three signals captured 73 percent of the trend – a satisfactory performance for a short-term trading system not specifically designed to follow trends.

The next example illustrates the ability of pattern-based systems to follow longer-term trends.

From September 1994 until August 1998, SPY was in an uptrend. The APS program was used to search for price patterns in SPY from January 1990 to Feb. 14, 2003. Patterns with a delay (i.e., the entry is delayed a certain number of bars after the initial pattern signal) of one to three bars were also considered. The search parameters were:

SPY
> 66%
> 28
< 4
open of next day
= 5% of entry price
one to three bars.

The results of the search are shown in Figure 7. For simplicity's sake, only the three patterns highlighted were chosen to build a trading system. One could have selected any subset or the whole set of patterns found. The specific selection was made to illustrate the potential of "delay patterns." All have a one-bar delay (shown as "open1" under the column "Trade on" in Figure 7).

This search produced a group of acceptable patterns for trading SPY. Three of the patterns were selected to be combined and back-tested.

Figure 8 shows the trade signals of a trading system using the three delay SPY patterns selected, over the 1994-1998 period. Table 2 shows the back-testing results over the same period.

The patterns found in the search from Figure 7 captured much of the long-term gain while being in the market a fraction of the time as typical trend-following indicator..

Source: MetaStock

The combined patterns found in te Figure 7 search produced the following results in SPY.

Source: MetaStock

The system generated 18 winners and no losers. The net profit was $68.83 – about 95 percent of the buy-and-hold profit of $72.19. Even though the system was in the market only 70 percent of the time, this profit was possible because some entry signals were generated immediately after a price correction, and the entry price was lower than the previous trade's exit price.

This is another advantage of using short-term price patterns. In certain situations, the system can exceed (sometimes significantly) the buy-and-hold profit.

Short-term price patterns offer an indirect but effective way to capture short-term – and sometimes long-term – price moves. By designing a trading system with the appropriate number and type of price patterns, the capability to follow price trends and generate profitable trades increases.

At the same time, the trader enjoys the reduced exposure of short-term trading, as well as its ability to generate a steady stream of profits. By combining the best of two worlds – the high profit potential of trend following and the flexibility of short-term pattern trading – performance and efficiency can improve noticeably.