A Robust Method To Interpret Advance Decline Issues

Advance Decline Issues, by its very natural, is noisy. Its values can be jumpy from day to day making it hard to interpret the information offered by the market breadth data. I am going to demonstrate a neat way to utilize the data that overcome the weaknesses in classic advance decline issues based indicators.

This is part of the Market Breadth Primer series.

Understand That There Is A Rhythm

All markets move with a basic rhythm – contraction and expansion. This happens to all the components within the basket of advance decline issues we monitor. Hence, there is also a hidden rhythm in the advance decline issues data. The problem is that it is very difficult to see that visually.

For many people, the advance decline issues data simply look like noise themselves. Following is a chart of Emini S&P and the NYSE Advance Decline Issues illustrating the point.

Changing the display style does not help.

At this point, it is clear that to interpret advance decline issues data beyond their single day readings and implications, it is necessary to transform the data first.

Using Net Advance Decline Issues Is A Bad Idea

Net advance decline issues (Net AD) is the end of day reading of the difference between the number of advance issues and decline issues. It is the basis of many famous market breadth indicators like the Advance Decline Line, Advance Decline Spread and McClellan Oscillator. The reason net advance decline issues was used because it combines the advance issues and decline issues data into one simple number to work with. It was also the only choice in the past because of the limited computation power at the time that these market breadth indicators were invented.

Net advance decline issues is not a proper representation of the underlying breadth data. It tends to exaggerate the readings with larger absolute values. Here is a comparison of net advance decline issues against the advance issues percentage readings based on a fixed total of 2000 issues.

Notice that net advance decline issues readings cannot be compared properly from one reading to another. For example, one day where the net advance decline issues is –500 and the next day with reading of +500 does not tell you the number of advance issues traded has doubled. Yet it implies the strength of the market is. This problem is magnified when we are looking at long term history when the number of traded issues changing over time. These mathematical issues with net advance decline issues make it not suitable for the identification of patterns. In short, net advance decline issues are statistically unstable.

Using Advance Issues Percentage Is A Better Idea

The better way to combine advance decline issues data is using the advance issues percentage or simply rate of change based on advance issues only. The former one has the advantage of long term stability and normalization for better statistical and pattern analysis. The latter one is good for adaptive analysis as it is a direct relative measures from one reading to the next. In this article, I will focus on the use of advance issues percentage. Rate of change based on advance issues has to be dealt with separately.

Following chart shows the Emini S&P with the NYSE Advance Issues Percentage.

The advance issues percentage data still looks like noise at the first glance, isn’t it?

By adding some lines to the chart, it may change your perception at once.

There are trends in the advance issues percentage data. They are very noisy but the rhythm of the market is clearly presented in the data. The catch is how to figure out the rhythm in advance.

It is often quite easy to see the developing trends in the advance issues percentage. Once you can identify the trend, you will be able to anticipate potential change in trend. This is the quality that breadth indicators like advance decline line are missing.

For some people, seeing trends in the data is a natural instinct. For these traders, they can probably use the advance issues percentage values as-is.

For others, they need a way to quantify the rhythm.

Noise Level Correlates To The Contraction Expansion Rhythm

As stated in the beginning of the article, the main problem with advance decline issues data is how noisy it can be. The level of noise in the data, however, is predictable based on range contraction / expansion in the related indices. Using the example above, when Emini S&P is trading within a tight range, the advance decline issues readings would be jumping wildly. That in turns produces extreme readings in advance issues percentage. Some people misinterpret this as a sign of directional change, but that is not necessary the case.

To better capture potential change in trend based on advance issues percentage, a dynamic reference zone has to be established. The trend zone would represent the current trend in the advance issues percentage. This zone at times have to take into account the increased noise level so that false signals can be reduced.

Early detection of the change in trend can then be accomplished by either the change in direction of the zone itself, or, a signal line when it enters and leaves the zone.

This technique is just an adaptation of signal processing on very noisy data.

Visualization Of The Trend Zone Concept

It is much easier to understand what I just explained above with an example. Follow is the chart of Emini S&P with its advance issues percentage, the trend zone and the signal line.

The trend zone is a band around the 6 period simple moving average of the advance issues percentage line. The band is a multiple of the mean deviation (read Mean Deviation vs. Standard Deviation if you want to know why it is preferred) of the advance issues percentage.

The signal line is simply a faster moving average.

Notice the turning of the trend zone are on time and often leading by a bar or two. It confirms the leading nature of the advance decline issues data. The signal line is just a way to highlight the potential turning points mechanically for the generation of mechanical trading signals.

Statistical Stability Produces Usable Mechanical Signals

If you have spend any amount of time on market breadth studies, you would probably notice majority of the market breadth analysis are discretionary. You rarely see anyone produces mechanical trading models with the breadth indicators. The reason is simple – it is not possible to produce stable trading models when the proper data pre-processing techniques are not employed.

Spending countless hours to apply standard indicators from the charting platforms hoping to find a working model is not going to help because market breadth data are not price data and they have different characteristics from price data. Market breadth data are better handled by signal processing techniques. This is one of those occasions where classic chart reading techniques are not going to help.

We can derive a trading system directly from the trend zone concept. Following is the performance chart of a straight forward implementation of the trading signals mentioned in last section. The line at the bottom of the chart is the net gain in dollars.

This model is always in the market, trades a single lot Emini S&P with commission, slippage, etc. all taken into account.

One thing making this model special is that it does not depend on price data at all. You can throw SPY in there you will find that it works well too.

This simple model has its ups and downs throughout the past 15 years. It is not perfect but it demonstrates the breadth analysis technique presented here has a clear statistical edge that one can lean on.

Share