Market-specific systems work best on a particular market because they capture some unusual feature of that market. It is difficult to speculate why certain markets show signature patterns. We should take extra care when developing such systems because the market mechanics driving such patterns could change abruptly.
The S&P-500 futures contract can be used to illustrate a pattern-based approach. For instance, we consider a continuous contract from April 21, 1982, through July 10, 1995, and test the standard simple moving average crossover system with 10-day and 11-day simple moving averages. We use a relatively loose $2,000 initial stop, which will absorb random price fluctuations, and allow $100 for slippage and commissions.
The 10- and 11-day dual crossover system lost $181,005 on paper, with 530 trades. Only 34 percent or 178 trades, were profitable, with a maximum intraday drawdown of $189,370. One interesting feature was that virtually all the loss ($185,545) was on short trades. This makes sense if we recognize that the market has been generally moving up since 1982. However, it is striking that this simple trend-following system fared poorly in spite of the prolonged uptrend. So the S&P-500 futures market is not a trend-follower’s delight.
Because all of the losses were on the short side in the previous test, it makes sense to try the simple moving crossover system in the antitrend mode. The antitrend rules are as follows:
- Buy if the 10-day SMA crosses below the 11-day SMA on the close.
- Sell if the 10-day SMA crosses above the 11-day SMA on the close.
Using the same test period, initial stop, and allowance for slippage and commissions as the previous test, the turnaround in profits with the antitrend rules was remarkable. This antitrend 10- and 11-dayJystem netted $55,920 for a swing of $240,925 on 531 trades. Fully 48 percent, or 254 trades, were profitable, with a maximum intraday drawdown of $32,735.
The results of the antitrend approach are not spectacular. However, they do highlight the unusual nature of the S&P-500 market. They suggest that you could find market-specific systems that would test poorly on other markets. For example, the 10/11 antitrend strategy lost $56,775 when tested on the Swiss franc continuous contract over the same period, but the 10/11 trend-following strategy lost just $13,088 over the same period.
The following is a glaring example of how “hindsight” influences system design. There were many “V” bottoms on the daily bar-charts of the S&P-500 market, so a bottom-fishing strategy that tries to pick bottoms was attempted. Theoretically, it should test well since this is an and trend approach. The rules for the S&P-500 “bottom fishing” pattern are as follows:
- A 20-day low has formed within the last 5 days.
- Today’s high-low range > X; X = 4 for conservative trades; X = 1 for aggressive trades (each point is not one tick, but one full S&P index point = $500)
- Today’s closing-opening range > Y ; Y = 3 for conservative trades; Y = 0 for aggressive trades.
- If rules 1,2, and 3 are true, then buy tomorrow on the close.
- Exit on the close of the twentieth day in the trade.
- Initial money management stop = $2,000 per contract.
Note that we can fully automate the bottom-fishing pattern. We have no difficulty getting entries, because if we get a signal today, we can buy on tomorrow’s close. So it is easy to implement using a mechanical system. For example, the analysis can be done after market hours, and the order entered before trading begins.
This system has a conservative entry combination and an aggressive entry combination. The conservative approach generates fewer trades. You can modify this pattern in many ways. The most obvious change is the exit strategy. For example, you could set an exit target at the most recent 20-day high.
The system was tested using System Writer Plus™ and actual S&P-500 contracts. The rollover date was the twentieth day of the month before expiration. The results are in two blocks in Table 4.20 because System Writer can process only 30 contracts at a time. You can treat either the conservative or the aggressive set of X and Y values as an unoptimized set. Both combinations were profitable on both blocks of data.
The equity curves for both options are shown in Figures 4.38 and 4.39. The equity curve for the conservative option is smoother than the aggressive option. Also, the aggressive option can produce larger drawdowns than the conservative values.
Data using the March, 1995 S&P-500 contract yield Figure 4.40, page 136, for X = 4 and Y = 3, and Figure 4.41 is for X = 1 and Y = 0. This system picked off the bottoms very accurately. Entry and reduced slippage are assured by entering and exiting on the close. Thus, a pattern-based, antitrend, bottom-fishing approach works nicely on the S&P-500 market.
You can try a variety of exit strategies. Instead of an exit on the close of the twentieth day (case 1), use a trailing stop on the 5-day low after a $1,000 profit on the trade (case 2). Case 2 with X = 4, Y •= 3, a $2,000 initial stop, and $100 for slippage and commissions from February 12, 1988, through July 10, 1995, had a profit of $59,025 over 44 trades (45 percent winners) with a drawdown of-$7,625. You can compare these data to the second row in Table 4.20 (case 1). Thus, the new exit strategy produced approximately the same profits, but with a smaller drawdown and more winners. The equity curves for case 1 and case 2 are shown in Figure 4.42. You can see that case 2 has shallower drawdowns than case 1.
To check the basic validity of the bottom-fishing pattern on other markets, we must modify the pattern slightly to make it more general. Values of X= 0.1 and Y = 0 are chosen in order to test across many markets. A trend-following exit, at the lowest low of the last 20 days, was chosen because not all markets are as dynamic as the S&P-500 market. The entry is switched to above the high of the signal day, instead of buying at the next days close, to reduce the number of entries in downtrends. The initial money management stop is $2,000, and as usual, $100 is deducted for slippage and commissions. The pattern uses all available data from January 1975 through July 1995 using continuous contracts on 17 markets. The results are for trading one contract at a time.
The generalized bottom-fishing pattern was profitable on 11 of 17 markets, including deutsche mark, Eurodollar, gold, Japanese yen, coffee, orange juice, Swiss franc, S&P-500, silver, 10-year T-notes, and the U.S. bond market. Thus the pattern also seems to work on markets that trend well or have good swing moves. The results are given in Table 4.21.
These data suggest that the bottom-fishing approach captures a basic trading pattern in the markets. The long test period and the profits on a variety of markets indicate that the idea is robust. The difference in performance between markets seems to be the amplitude of the movement after forming the pattern.
An extension of the test of the bottom-fishing pattern to stocks explores its performance over different time periods. Figures 4.43 (weekly) and 4.44 (monthly) illustrate how the generic bottom-fishing pattern works. Figure 4.43 has weekly data for Union Carbide showing how the pattern picked the bottoms in 1990 and 1991. The pattern also stayed long throughout the major uptrend. The pattern tests well with weekly data on stocks. Figure 4.44, page 140, has monthly data for Caterpillar Tractor. The bottom-fishing pattern responded to the 1992 bottom and stayed with the stock throughout the rally.
In summary, the bottom-fishing pattern-based system is a good example of a market-specific system. You can use it as a model to develop other pattern-based systems on the S&P-500 market. The pattern can be generalized successfully to other markets, including stocks. The bottom-fishing pattern also works across time periods such as daily, weekly, or monthly. Thus, the bottom-fishing pattern captures a fundamental pattern of price evolution.