Testing the Pattern : Types of Market Test

ADDING A VOLUME FILTER

Let’s look at requiring a volume filter in the basic hanging man pattern. The first volume filter to be tested required that the volume on the day the hanging man occurred be larger than the previous day’s volume; yet, after testing, this requirement resulted in fewer trades than would be expected, but yielded about the same results. Therefore, when defined as above, this filter does not appear to be something useful.

Traders make their living by examining various trading patterns looking for a combination of patterns and filters that provide a statistical edge. Finding filters that do not work is helpful because the trader knows those characteristics may be ignored when looking at trading candidates.

Using a filter that requires the volume on the day the hanging man occurs to be above the 21 day simple moving average of the volume shows some improvement. The loss is cut in half and the winning percentage improves slightly. A loss is still a loss, so even though this filter results in a smaller loss, it would be better for the trader to not use the basic hanging man with this filter. Using multiples of 110% and 120% of the average volume did not significantly improve results either.

Filters Are Not One Size Fits All

It appears that large volume on the day the hanging man occurs is not an effective filter. Some patterns, such as certain types of pullback patterns, show improved results when using increased volume filters. This test data indicates that just because a filter helps one type of trading pattern, it does not mean it will help others. Traders need to be careful about applying filters that are recommended by others. I always want to test any idea to see how it has performed in the past before using it.

TESTING THE PATTERN UNDER DIFFERENT MARKET CONDITIONS

Most trading patterns have market conditions in which they work well and other market conditions in which they do not work well at all. The initial test period of 01/03/06 to 05/01/ 07 shows bullish and bearish environments as shown in Figure 5.6.

FIGURE 5.6: MARKET ENVIRONMENT DURING 1/03/06 TO 5/01/07

BEAR MARKET TEST

The period between 05/05/06 and 07/21/06 was bearish and saw the NASDAQ dropping from the 2,350 area to the 2,015 area. During this period, the backtesting results showed 270 hanging man patterns with more than 79% of them resulting in profitable trades, as shown in Figure 5.7.

FIGURE 5.7: HANGING MAN TEST RESULTS DURING 5/05/06 TO 7/21/06

This is a very interesting result. In this particular bearish market, the hanging man was much more effective than in any of the previous tests that encompassed both bullish and bearish periods. The annualized ROI for the buy and hold strategy showed a loss, while the hanging man showed a significant gain. As noted above, the test results for any single period may or may not be indicative of actual results, so further testing of the effectiveness of the hanging man pattern in bearish markets is required.

Testing the hanging man pattern during the market’s bearish period of 07/28/05 to 10/17/05 showed good results with more than 59% winning trades and an annualized ROI of nearly 33% as compared to negative 16% for a buy and hold strategy during the same period. The NASDAQ was also in a clear downtrend during the five month period of 12/30/04 to 04/28/05, as shown in Figure 5.8. Test results indicate that during this time there were 1,032 hanging man patterns and taking all the trades would have resulted in nearly 57% winning trades. The annualized ROI for this period was 30%, as opposed to a negative 17% for buy and hold as shown in Figure 5.9.

FIGURE 5.8: NASADQ DOWNTREND OF JAN. '04 TO MAY '05

FIGURE 5.9: HANGING MAN RESULTS DURING BEAR MARKET OF JAN. ‘04 TO MAY ‘05

Testing the hanging man pattern has shown that it performs poorly in periods when the market is moving through both bullish and bearish phases, and that it performs well in three different bearish environments. This does not prove that it will work well in all bearish market environments, but it does indicate that it is worth looking at in bearish markets.

FAILURE IN THE BULL MARKET

The initial test period of 01/03/06 to 05/01/07 for which the test results were negative, as shown in Figure 5.4, contained a bullish market period between 08/11/06 and 11/17/06. This bullish market period is shown in Figure 5.10, and the test results for this bullish period are shown in Figure 5.11. During this bullish period in the NASDAQ, the hanging man pattern only showed 41% winning trades and an annualized ROI of negative 28%, while buy and hold showed a positive 36% annualized ROI.

FIGURE 5.10: BULLISH MARKET PERIOD OF AUG. 2006 TO NOV. 2006

FIGURE 5.11: HANGING MAN TESTS RESULTS DURING BULL RUN OF AUG .2006 TO NOV .2006

The market was also in a bullish period between 10/13/05 and 12/02/05. During this period, the test results indicated the hanging man pattern was only profitable on 34% of the trades and lost an annualized ROI of 55% as compared to a 52% annualized ROI gain for buy and hold. The hanging man continued to show good results in bearish markets, poor results in bullish markets, and mixed results in mixed markets. Based on these results, I would not trade the basic hanging man pattern in bullish environments and would consider it one of the tools in my toolbox during bearish markets.