|The use of time rather than price has largely been attributed to legendary chartist W.D. Gann. Noted traders such as Arthur Merrill and Larry Williams have used seasonal patterns to trade both stocks and stock index futures, and Stock Trader's Almanac has done a good job of identifying seasonal trends. It is clear, however, that merely using time as a trading tool may not lead to optimal outcomes. Using price to filter seasonal trends may have more potential, and it is widely believed that many large institutions use price action to utilize seasonal trends. |
The first week of the trading month contains exploitable seasonal tendencies. The aim is to develop a long-only model based on price action to trade this seasonal trend. The primary goal was to develop a model capable of trading the emini Russell 2000 futures.
Several exits are needed to accomplish different goals:
We chose the system trading program BEHOLD! because we were familiar with its coding, its plethora of built-in time-based functions, and coding efficacy. The BEHOLD! code is provided here:
if wom>3 and ran
FIGURE 1: THE BEHOLD! SYSTEM. Here you see the buy & sell signals on the emini Russell continuous futures contract. A commission of $7 with slippage of $20 was used with a margin of $4,000. A big point value is $100 for the emini Russell.
The performance summary table in Figure 2 indicates average trades of $350 with a 52% win rate and a profit factor of 1.72. The average winner lasted nine days, with the average loser lasting only five days. More important, the maximum loser was only $2,455 with a maximum closed drawdown under $10,000. The model only trades about 10 times per year. The trend filter kept the model from trading, for example, from August to late December 2008. The emini Russell futures went from about 715 to about 430 during this period. The equity curve is shown in Figure 3 and illustrates the lack of major drawdowns.
FIGURE 2: PERFORMANCE SUMMARY OF THE EMINI RUSSELL FUTURES (DECEMBER 11, 2001-FEBRUARY 9, 2011). Here you see that average trades are $350 with a 52% win rate and a profit factor of 1.72. The average winner lasted nine days and the average loser lasted only five days. More important, the maximum loser was only $2,455 with a maximum closed drawdown under $10,000.
FIGURE 3: EQUITY CURVE OF THE EMINI RUSSELL FUTURES (DECEMBER 11, 2001-FEBRUARY 9, 2011). As you can see, there are very few drawdowns.
The emini Russell data reflects recent information to assess how this model would compare over a much longer term; we tested it on the cash Standard & Poor's 500 between 1971 and 2010, with the gain and loss set to percent rather than cash values.
Over 256 trades, the model proved profitable, although the holding periods were longer and accuracy slightly lower. The maximum drawdown was 28%. The model mostly avoided trading during the 1987 and 1990 market declines. The performance summary is displayed in Figure 4 and the equity curve is shown in Figure 5.
FIGURE 4: PERFORMANCE SUMMARY OF S&P 500 CASH (1971-2010). Over 256 trades, the model proved profitable, although the holding periods were longer and accuracy slightly lower. The maximum drawdown was 28%.
FIGURE 5: EQUITY CURVE OF THE S&P 500 CASH (1971-2010). Over this time period, the system proves to be very profitable.
Using the model on the big five stock index futures may increase the number of annual trades, although diversification may be an issue. However, the availability of emini contracts would work well in this regard, and the user could diversify among the five index futures for the price of one big contract. The success achieved in the Russell could be reproduced in the other major indexes.
We were pleased with its utility in trading the EFA (iShares MSCI EAFE index fund). Moreover, the basic premise appears to hold true not only for recent data but also for data going back several decades. While we do not suggest that readers trade this in its current form since our intent was merely to show how to combine price action with seasonality, they may wish to explore improving the model. These modifications could include using an opening range breakout rather than exiting on close as well as exiting based on a range function rather than percentage.
Combining seasonality with price action can be a worthwhile addition to the trader's arsenal of trading systems. Other seasonal trading patterns are well discussed by B. Thackray and B. Lindsay and offer the potential for similar price action modifications.