It appears once the NASDAQ Composite (COMPQ) Advance/Decline Ratio moves to extremes, a temporary trend reversal occurs. When trying to use this breadth tool as a contrarian indicator to capture this reversal, the trader is confronted with defining an extreme level. Given potential changes to the underlying index and changing market conditions, what constitutes an extreme A/D Ratio reading may very well be a moving target.

This article tests one possible approach for the moving target dilemma. The COMPQ A/D Ratio is used to signal trades for QQQQ, the NASDAQ-100 (NDX) exchange traded fund [ETF]. See last week's article ("Relative Advance-Decline Levels," June 18, 2009) to review the correlation of returns for the index versus the ETF, along with a chart that displays A/D Ratio movement with QQQQ.

Adaptive Indicator Levels

There are a variety of ways a trader can create and adaptive technique. Perry Kaufman's book, Trading Systems and Methods, describes his adaptive moving average approach along with others. The system tested here uses a fixed, simple moving average (slight oxymoron here), meaning there are no changes to the moving average setting. It also uses the same period of time to calculate a rolling standard deviation [SD] so the extreme value varies over time. If the two lines were viewed on a chart with the A/D Ratio, you'd have the ratio values and a simple moving average [SMA] with a lower Bollinger Band.

Edge

The system seeks to benefit from a QQQQ bounce after extremely low A/D Ratio readings. A buy signal is generated when the A/D Ratio is less than its 150-day SMA minus 0.75 times the 150-day SD. The SD multiple was not optimized, but it did provide the best results for different settings tested (0.5, 0.75, 1.0, 1.5, 1.75 and 2.0).

To obtain sufficiently extreme values for the A/D Ratio, a 150-dy SMA and a 150-SD was used to trigger a signal. The value that constitutes an extreme then changes over time and reflects recent market conditions. One problem with the indicator used in this approach is that each day there are two impacts to the new signal values:

1. The most recent ratio data added typically changes the indicator value and
2. The oldest data point that drops off typically changes the indicator value.

One way to help minimize this impact is to use an exponential MA instead of a simple MA.

This approach was tested since there were consistently positive returns noted for basic breadth and sentiment systems in Robert Colby's, The Encyclopedia of technical Indicators. The data was downloaded using Worden's TC2007 software.

Results

The A/D Ratio signals a buy at the open on the following trading day when its reading is below the 150-dy SMA - (0.75 x 150-dy SD). The position is closed two trading days later, at the close. For instance, a signal generated on 9/8/2005 would result in trade entry at the open on 9/9/2005 and trade exit at the close on 9/13/2005.

Average returns were small over the three year period (9/1/05 - 8/31/08), but the position was only held for a total of three market days (two overnight periods). There was also an average of 36 trades per year which can allow for returns to compound. Although more appealing without commission and slippage costs, the system maintained profitability when costs of $21 per trade were added. Table 1 provides the results with and without costs.

Table 1: A/D Ratio QQQQ Bounce System Results With & Without Costs

Since returns can be compounded the system is expected to provide positive returns over time. Expectancy is calculated as follows:

Expectancy = (%Win x Avg Win) + (%Loss x Avg Loss)

While profitable in a back-test period, systems can degrade over time. There is a concern about negative expectancy going forward since the returns are small. Ideally the adaptive technique will help minimize this.

A Contrarian Edge from the A/D Ratio

Extreme values for different breadth measures may be used as a contrarian tool. Given the changing nature of exchanges and markets in general, breadth readings can also change. Rather than using a specific value to denote market extremes, traders may want to consider using a representative average +/- some number of SDs to identify extremes. The question is: how do you decide when to update your readings?

There are a few different approaches traders can use to update extreme readings for indicators. Ideally the approach will be objective while still remaining in tune with current market characteristics. The first test looked at creating a long position in QQQQ at the open the day after an extreme low was reached (low total advancers and/or high total decliners). The test data used NASDAQ A/D Ratio values below 0.308 as a trade signal, which is close to the average minus 1SD value using all data (0.332).

A timed exit was implemented to test the system with returns evaluated using the closing value one day, two days and three days after the position was established. Although the average return was modest, the short-term holding period (lower risk), multiple signals throughout the year and compounding of returns, merited a further look.

An adaptive, objective technique for setting the A/D Ratio trigger level was necessary for me to feel comfortable with the approach. This was accomplished using a trigger based on a 150-day moving average for the ratio less some multiple of the 150-day standard deviation [SD] of the ratio. Although the system was not optimized, it did appear to work best at 0.75 the SD in a three year back-test (110 trades). Using data from 1/28/2005 forward to calculate the values, a buy signal is triggered as follows:

Buy QQQQ, when A/D Ratio today < 150-dy SMA of Ratio - (0.75 x 150-dy SD of Ratio)

The indicator looks like a Bollinger Band with only one lower band rather than an SMA with two bands on either side. Readings below the band's value triggers an entry the next trading day.

To access other articles written by Clare White, please click here.

Clare White
Contributing Writer and Options Strategist
Optionetics.com ~ Your Options Education Site
Questions for Clare? Visit the Optionetics.com Discussion Board