Confessions of a Market Timer
By Robert Fischer

A fund manager whose mode is active money management and chart analysis gives serious thought to what a successful trading system ought to look like. He may or may not prove to have the Holy Grail, but he has some useful thoughts to contribute to the quest.

If the professors in the preceding article had an academic interest in price predictability in the stock market, Robert Fischer has a practical one. He has been looking for market-beating strategies for a long time. Back in 1980 he was a proponent of the Merton and Scholes technique of maintaining a portfolio allocated 90% to Treasurys and 10% to selected stock options. The results in terms of return and risk, he noted, seemed to be better than, say, the Dow Jones Industrial Average in rising markets, worse than the Dow in stagnating markets, and better again in falling markets. He found it a solid yet simple strategy: Work could be concentrated on just two days each half-year, when existing six-month calls had to be sold and new six-month calls had to be bought.

But that was not enough. Fischer has long searched for active-management strategies that could outperform index funds, whether with higher returns or lower risk (or, ideally, both). He was also a believer in technical analysis; his particular denomination was adherence to Elliott (as in Elliott Wave Theory) and Gann (William D. Gann, a now little renowned theorist active mainly in the 1930s and 1940s, when he had a fairly wide following).

In recent years such a quest has seemed even more urgent for active money managers like Fischer, when stock-market indexes of all types have typically returned 15% to 20% a year, and the S&P 500 has returned substantially more than that: Indexes have been tough to beat. Another reason for pursuing such a search: The rising importance of emerging markets, where fundamental information can be incomplete, difficult to get, and often just plain wrong.

When undertaking this quest, there were two major issues to think through: What should one look for when searching for such a strategy? And, realistically, how good can it get?

Obviously, the successful active manager had to anticipate market moves well enough. Fischer concluded he needed a set of clearcut, objective rules or principles - ones that work. They should offer an especial advantage because they presumably would give unambiguous signals to alert him to and minimize bear market conditions and maximize bull market opportunity. But such rules can be very difficult to apply if they are not clearcut and objective, especially in market environments characterized by strong, erratic price moves. The manager can use something definitive to the point of automaticity to steady his nerves at times like that.

Soon, three main requirements seemed to arise, given Fischer's preference for technical analysis: automation and computerization; a model that produces a definitive answer; and what eventually came to seem most important, a process of dynamic asset reallocation in the face of changing market circumstances.

1. AUTOMATION AND COMPUTERIZATION

Computers could provide two essential functions. One, they could enable the manager to handle enormous amounts of historical and other data. Two, they could apply a set of well-defined buy or sell signals and/or stock selection rules nervelessly and unemotionally, unaffected by the excursions and alarms of a disconcerted marketplace.

One trap of working with computers to analyze past data needed to be avoided, however. Fischer could create mathematical, computerized models that perfectly fit all the historical data (see preceding article), but who is to say that the model will work on new, i.e. future price movements? And if it does work for a while, if market prices continue to move within the channel of historically determined parameters, will the model still perform when market conditions or economic conditions inevitably change and push investors out of the familiar channels?

What was needed, it seemed, was a shift in perspective away from seeking the best fit to the past and toward a strategy of seeking the best average fit to many different scenarios and cases, to many different markets and conditions, and to as many different securities and investment products as possible.

2. ONE-PARAMETER APPROACH

Active managers are quick to blame unpredictable changes, like the recent turmoil in Asian markets, for example, for losses they take. But the investing shoe should be on the other foot, Fischer argued. Since the unexpected does happen - a circumstance painfully experienced last summer by the hedge fund Long-Term Capital - such a trading model should be designed in such a way that even extreme market situations will not hurt performance too badly. Thus, the same set of rules ought to operate constructively in virtually all market situations.

Fischer's longstanding fascination with the theorizing of Elliott and Gann led him toward seeking a solution in the area of pattern recognition applied to price series. Since charting attempts to identify patterns of change in investor sentiment rather than underlying economic circumstances of individual securities, this approach could possibly yield his universal model, since all securities are bought and sold based on the decisions - and sentiment - of investors.

Fibonacci ratios (a class of numerical sequences familiar to mathematicians since the 12th Century); Gann speed, fans and angles; and Elliott Waves were thus applied to price data in earnest. "Gann's proportional relationship between time and price is the basis for his theory of geometric angles...Gann's geometric angles are trendlines drawn from prominent tops or bottoms at certain specific angles. Those angles are determined by the relationship between price and time...Abull market is in force as long as prices are above [a] rising line [drawn from a bottom]. Abear market is in force as long as prices remain below [a] declining line [drawn from a top]. When prices in an uptrend decline [to the trendline], time and price are in perfect balance and a state of equilibrium exists. The breaking of the trendline therefore indicates a shift in that relationship and a possible change in the trend. Channel lines can also be drawn from prominent highs and lows that are parallel to the basic trendline." [Descriptions from several published sources; see References].

Obviously, familiar chartist ground. But the rising bull trendline, if used as a sell, or stop-loss, signal, would tend to protect profits. The declining bear trendline would help get an investor back in at more favorable levels. In short, they might create a favorable investment bias. As Fischer tinkered with his formulas, he reports he made an interesting discovery. He developed a random price generator by which price moves in all kinds of trending products could be modeled. Applying his evolving model to these - random - data brought results he found remarkably close to results achieved on the real historical data. This test suggested that the buy-sell formulas being developed may have useful investment application whether or not the method was strictly speaking predictive of future results at all.

3. DYNAMIC ASSET ALLOCATION

Analysts frequently argue about the best combination of cash, bonds and stock in a portfolio for controlling risk. But if the need is to adjust to changing, unexpected market conditions, being 20%, 30%, or 40% in cash, say, in any given market should not be the key question, Fischer concluded. What would be needed is a model more or less automatically adjusting the portfolio anywhere from 100% stocks in a strong uptrend environment to, ideally speaking, 100% cash (or short) during strong downtrends, and perhaps somewhere in between as risks seem to change or markets turned uncertain.

This allocation process, too, would need to be conducted according to clearcut, established rules, to reduce or eliminate subjective decisions and the temptation to ignore the signal being given because "it's different this time" - an action almost every investor has fallen into, and often to his or her regret.

IS THERE SUCH A TRADABLE SYSTEM?

Fischer developed a two-step system: one for timing buy and sell actions, and one for selecting stocks or mutual funds for a portfolio.

What Fischer came up with and presumably is using is proprietary, of course, and Indexes is not in a position to test or verify his results independently. However, what follows is the results Fischer reports from following the line of logic and experimentation described above:

While testing his method real-time for three years on a collection of stock index futures contracts, bond futures, and cash currencies, and getting respectable net returns of 20% in 1996, 29% in 1997 and 27% in 1998, he set up a small model portfolio of 40 mutual funds to be traded real-time and a larger model portfolio of 100 international stocks to be paper-traded.

The selection of the 40 funds and 100 stocks was done by a momentum indicator analyzing weekly data from Datastream International from a universe of roughly 3,000 stocks and 1,000 mutual funds.

For 1998, he reports a 17% return for the 40 mutual funds and 21% for the 100 stocks, and that his model increased and decreased the portfolio cash positions roughly in line with the fluctuations of the stock market, as set out in the following table:

For first quarter 1999 he reports gains of 4% and 10% respectively. Fischer's tests were global in nature, but there was nothing to keep him from applying his technique to sub-universes of the investor's choice - by region or country, or by industry group.

TAKING ON THE S&P 500

Fischer also backtested his method on the S&P 500 for all of the 1990s through 1998. He applied his timing model to the entire index (labeled "500" in the table below), and then constructed a model portfolio of 50 stocks (labeled "50") using his momentum method for an additional test. Here is what he reports:

Many active managers have found the decade of the 1990s, and particularly the last few years of it, to be a bad time to be out of the market for any significant period, and Fischer's pure timing model had the same experience - although it beat the S&P in the latter's four weakest years and showed visibly less volatility. Overall, Fischer figures it captured two thirds of the profit potential at half the risk, if the latter is measured by degree of retracement in the valleys between price peaks.

On the other hand, applying a momentum strategy as well, he reports, resulted in a spectacular series of gains.

He reports a different result if he adds a bit of judicious short-selling, as called for in his method, to the options of being long or in cash.

Here is the result Fischer reports from a backtest using a strategy of being either long or in cash:

In each case his measure of risk was substantially lower than the same measure for the S&P500. Note that the addition of short selling did little for actual return - this has not been a good decade for bears - but that it reduced his measure of volatility. It might have a bigger impact in a future bear market, a situation in which his technique has not had a chance to be tested.

It is worth noting that if Fischer's method proves out, it would apply to international markets where fundamental analysis is inevitably incomplete or risky because of more or less great lack of transparency in more than a few of the world's securities markets. In fact, here are the figures he presents comparing his model 100-stock (paper) portfolio of international equities to Salomon Brothers' benchmark bond index (B), Morgan Stanley's EAFE (E), and Morgan Stanley's World index (W):

Whether Fischer has grasped the Holy Grail or not remains to be seen: It would have to prove reproducible by others. But the general framework of his thinking as he reports it would seem to merit the thought and attention of anyone, true-believing chartist or no, who wants to undertake market-strategy research along similar lines and see where it leads.

REFERENCES

A Simulation of the Returns and Risk of Alternative Option Portfolio Investment Strategies; Robert C. Merton, Myron S. Scholes, Matthew L. Gladstein; unpublished working paper, 1976.

Fibonacci Applications and Strategies for Traders, Robert Fischer; John Wiley 1993

Technical Analysis of the Futures Markets, John J. Murphy; The New York Institute of Finance, 1986

Handbook of Futures Markets, Perry J. Kaufman; John Wiley, 1984

The Encyclopedia of Technical Indicators, Robert W. Colby & Thomas A. Meyers; Dow Jones-Irwin, 1988

Volume Cycles in the Stock Market, Richard W. Arms, Jr.; Dow Jones-Irwin, 1983