Building a viable indicator is only half the battle. The real trick is in applying that indicator profitably to the markets. The history of stock market trading is filled with forgotten treatises that designed analysis tools that worked great on paper but — whether because of logistics, faulty premises or the failed logic of their practitioners — blew up marvelously when used in real time in real markets.
In the first part of this series, we discussed advance-decline (A-D) analysis and showed how an A-D line based on the 20 Most Actives (MAAD) theoretically is superior to the more popular New York Stock Exchange A-D (NYAD) line. We explained the development of MAAD and identified key considerations and assumptions necessary in its construction. We also showed how it performed anecdotally.
Here, we will extend our assessment of its application by developing a system based on MAAD and testing it.
After weekly Most Actives and NYSE A-D data are collected and calculated, an algorithm extracts the highest and lowest monthly values from MAAD and NYAD. Exponential moving averages of “n” bars are applied to the highs and lows of the S&P 500, MAAD and NYAD to create moving average channels. A Slow K stochastics indicator based on MAAD or NYAD confirms buy and sell signals. The strategy rules are as follows:
- To generate a buy signal, the monthly high of the S&P 500 must rise above the upper line of the exponential moving average channel of the S&P 500 highs; the monthly high of MAAD or NYAD must rise above the upper line of the exponential moving average channel highs of MAAD or NYAD; and Slow K applied to MAAD or NYAD must rise above 0.50.
- To generate a sell signal, the monthly low of the S&P 500 must decline below the lower line of the exponential moving average channel of the S&P 500 lows; the monthly low of MAAD or NYAD must decline below the lower line of the exponential Moving Average Channel lows of MAAD or NYAD; and Slow K applied to MAAD or NYAD must decline below 0.50.
The Tradestation code for the above system is shown in “Coding the system” (below).
After the initial data computations and observations of MAAD and NYAD charts, the assumption was that MAAD ought to consistently outperform NYAD over time. The objective algorithm constructed and tested underscored that original conclusion (see “MAAD system performance,” below).
Despite the insertion of a wide variety of duplicate inputs in the MAAD and NYAD algorithms, in no instance did NYAD outperform MAAD. NYAD consistently demonstrated lower levels of profitability in all instances while also creating more trades, many of which were unprofitable. Profit ratios also were noticeably lower in NYAD, as were the cumulative compounded percentage returns and compounded annual percentage returns.
Analysis in context
The 50-plus years of data included in this article reveal three general periods of stock market activity.
The first phase was the broad stock market “consolidation” beginning in the mid-1960s that lasted until index prices broke out to new all-time highs: The Nasdaq Composite index in September 1978, S&P 500 index in July 1980 and the Dow Jones Industrial Average in November 1982. MAAD held substantially above its June 1962 bottom into the 1974 lows to suggest accumulation. But NYAD made a lower-low in 1974 to erroneously indicate distribution and actually was more bearish than index pricing going into the major low in 1974. MAAD broke to new highs in December 1976 following its May 1969 high, while it took NYAD more than 17 years, until June 1983, to better its January 1966 peak.
The second phase was bullish, persisted until early 2000 and includes the primary bull market that lasted from October 1974 until March 2000. MAAD not only predicted that uptrend by highlighting big money accumulation preceding the 1974 lows, but it also generated a long-term sell signal in October 2000 at the onset of a two-year bear market. During that 26-year advance from 1974 to 2000, MAAD remained consistently bullish and was not shaken out despite market weakness in 1984, 1987 and 1990. By contrast, NYAD predicted a market top in April 1998, nearly two years too early.
The third phase that began in 2000 includes two bear markets, in 2000-02 and 2007-09, and two bull markets. During the first bear market, NYAD prematurely was bullish two years before the bear market lows of October 2002. The first bull trend lasted from 2002 to 2007, while the second, begun in March 2009, yet has to be fully resolved.
While our trading algorithm forgave NYAD the drawdown from 2000 to 2002 to the extent the S&P 500 eventually recovered, not many investors would have been able to suffer through the ensuing 64% drawdown in equity during that period. MAAD suffered no such drawdown because it remained in sync with the market.
The equity curves for each indicator are shown in “Equity curve analysis” (below).
Of course, with any trading strategy or methodology, there are a number of criticisms that can be levied.
Because the trading results for MAAD ultimately rely on the net up/down count of the top 20 Most Actives, the maximum fluctuation in MAAD necessarily is capped at 20 issues, even though there are no such limitations on price movement or volume. An example of this phenomenon can be found in the October 1987 Crash, when MAAD apparently did not reflect the unusual magnitude of the price decline. In this example, however, MAAD correctly identified that decline as potentially no more than an intermediate-term pullback within the context of a major cycle advance. This indication was in sharp contrast to widespread fear at the time that the Crash of ’87 was the beginning of a new bear market and an economic depression.
In addition, it could be asserted that the 20-fold increase in exchange volume over the past 50 years could have made it increasingly difficult for MAAD to adapt. To the contrary, MAAD has adjusted well because volume, as such, is less important to MAAD than the direction of the top 20 volume leaders insofar as the MAAD net issues count is concerned.
Another criticism of the procedure could be that collecting Most Actives data on a dollar volume basis makes more sense than using share volume. For example, on Dec. 9, 2011, Bank of America, quoted at 5.72 on 1.3 billion shares, traded more share volume than the S&P SPDR exchange-traded fund (ETF) quoted at 126.05 on 1.09 billion shares, yet the dollar volume was only $7.525 billion for BAC versus $108.097 billion for SPY. Such differences average out, however, because the majority of Most Actives issues tend to move in the same direction at the same time. As more dollar volume data become available in the future, we will be able to gain greater insights into the share volume vs. dollar volume question.
A better measure
For decades, the advance/decline line has been used to measure market breadth and to identify trends and trend divergences. In this two-part series, we sought to demonstrate that conventional A-D data, such as those reported by the New York Stock Exchange, are seriously flawed both in terms of the raw data and the subsequent results derived from that data. But when 20 Most Actives data are inserted into an A-D line that is massaged in an objective strategy trading the S&P 500, the Most Actives outperformed the NYSE series on every quantifiable level.
There is no practical way to fix an A-D line that uses conventional stock exchange A-D data. The biggest fault of the indicator is that it relies on unweighted data that are used in a cumulative A-D line that is compared to a weighted index like the S&P 500. And the process of excluding data, such as bond funds, preferred issues, closed end mutual funds and ETFs in favor of pure equities in the hope of repairing the A-D line, only will create a diminishing pool of issues from which to measure market health. Volume must be the sole determinant of inclusion, not asset purity based on subjective or arbitrary rules.
When Most Actives data are substituted for flawed A-D data, that change not only improves trading results, but also gives analysts a better reflection of big money tendencies because that segment tends to be, in effect, “the market.”
This research was limited to Most Actives monthly data and long-term trends evident over the past 50 years. These findings should stimulate additional research into applications of the Most Actives on all cycles, including daily and weekly data, as those shorter time frames offer opportunities for more frequent data sampling that can be rich in analytical possibilities.