For many traders, the S&P 500 and the E-mini are their speculative vehicles of choice, with huge daily volume and open interest. These contracts present opportunities for day-trading profit and serve as a key tool for long-term equity investors and hedgers. The long-term investor is willing to overlook blips and corrections to achieve the big move when it occurs.
Long-term investors have different goals than short-term day-traders. The long-term investor may be interested in holding out for a big move, willing to overlook the blips and corrections along the way. These investors are typically satisfied with lower rates of success for the payoff of bigger profits when profits come. For these traders, the S&P 500 also offers significant opportunities.
The methods used to carve out those opportunities, however, may not be created equal. We will compare three models and see how much of the long-term moves we can catch. Our data set will be the closing price of the cash S&P 500 on a weekly basis. We will test a 52-week simple moving average model, a 105-week exponential moving average model and a price-action indicator with a one-year reference. Most charting vendors provide simple and exponential moving averages, so following the models should be simple.
The strategies are straightforward. On the first two, the market is sold if the current week’s close moves below the average and the market is bought if the current week’s close crosses above the average. The price-action indicator is computed by taking the current week’s close and subtracting from it the price one year earlier (52 weekly bars). A simple three-week moving average of these prices is used to generate signals.
We can reduce false signals with the same two restrictions on all three indicators. One, we require the reading to be at least one point higher or lower on the breakout (so, a reading of -0.79 would be insufficient) and two, any reversal within four weeks following the prior signal will be ignored because it is easy to have choppiness when these signals turn. Positions will be entered immediately on the positive signal and the trade will be maintained for four weeks regardless of a reversal signal.
Our data begins in 1950, but the three-point range of the first three years makes the effective start 1953. First, we equalize the three methods. They may begin trading when the first strategy signals a trade. That date and price are fixed as the start. In this case, it’s the close on Jan. 22, 1954, at 25.85
By way of a benchmark, the S&P gained a net of 1,076.57 points between 1954 and December 18, 2009 (see “Field of dreams”). Of course, it traded much higher in the interim reaching 1561.80 on a weekly closing basis for a maximum gain of 1,535.95 on the long side. Reversing to short at that price and holding to the present time would have brought the total gain possible from just two hypothetical trades to 1,968. But a steadfast buy and hold strategy from 1954 would have yielded 1,076 S&P points.
We require any successful system to exceed 1,076 points, and will judge a system a significant success if it achieves even 65% of our total ideal of 1,968 points, (i.e. 1,279 points before slippage). The fewer trades it takes to yield such points, the more successful the system will be, since every trade adds the potential for slippage and error.
A 52-week moving average should enable us to smooth out most of the chatter one might expect from a four- or eight-week moving average. Adding the filters we have imposed above also should help in theory. To our surprise, a simple 52-week moving average is a very poor system. It trades 80 times and only 45% of the trades are winners, including one open trade.The gross point gain is 1484.24 before slippage. If we add 0.30 points for slippage and commissions (i.e., $150), this reduces the gross gain to 1460.24 points. The model gave a buy signal on the close of July 24, 2009, booking a gain of 351 points from its sell signal on Jan. 4, 2008. Ironically, almost exactly 30% of the model’s gain was earned in the last closed trade. The best trade was a gain of 561 points from the trade between January 1995 and August 1998. This model, with 55% losing trades, was much worse than the accuracy level we desire, and we deem it a relative failure, although it did exceed the target of 1,279 points with an average gain of 18.25 points per trade.
The 105-week exponential moving average fared worse in accuracy and in net gain. It traded far fewer times: 26. There were 11 winners, or 42.3%. The gross point gain dropped to 1,369.84, including an open trade profit of 175.91 points. The net gain after slippage is 1,362.04. The average trade gained 52.38 points. Of the three models, the 105 EMA is the only one remaining short with a trade initiated on June 23, 2008, at 1278.38, and an open profit of 299.12. The EMA trades far less often and is less sensitive to chatter than the SMA, but also is slower to respond, in part due to its longer period. Almost all of the gain was earned in a single trade begun on Aug. 27, 1982, and culminating on March 16, 2001, for 1,033.42 points. The model stayed long for almost 19 years.
Finally, our third model is the one year indicator based upon a simple philosophy. If current price is better than one year ago, the market is improving and we want to be long. If the S&P is lower than one year ago, the market is declining and we want to be short. Again, we use a three-week average of the difference to try to eliminate any chatter and apply the same filters as in the previous two models and start on the same date. This model also had 12 winners, but only traded 19 times including an open profit of 34.47 points; 63.15% winning trades. The gross gain was 2,156.73 points and the net after slippage is 2,151.03. Notice that this exceeds the ideal hypothetical of 1,968, so we conclude this is a resounding success. The best gain was a long on Feb. 10, 1995, closed on Nov. 10, 2000 for a gain of 883.54 points. The average trade gained a resounding 113.21 points.
In the 105-week model, only three of the 28 trades (11%) had gains exceeding 100 points, including the open trade. In the one-year indicator, six of the 17 trades (35%) exceeded 100 points, including the current gain. Eight of the 80 simple moving average trades (10%) exceeded 100-point gains. On the losing side, the 52-week average had four losses in excess of 100 points. Neither the longer-term average (worst loss: 92 points), nor the one-year indicator (worst loss: 43 points) suffered a 100-point loss on a closed trade (see “Long-term Titan”).
The ideal trade is to have bought at the bottom and sold short at the top, always good advice in any market, because you only need to stop by to trade twice in a lifetime. That’s good work if you can swing it. That said, our models indicate the following:
a) Trading less may be better,
b) The longer-term exponential moving average is inferior to the shorter-term simple moving average,
c) The one-year indicator is significantly superior to both moving averages,
d) The one-year indicator is better than the ideal trade and may be of significant assistance to the position trader
An investor or a long-term trader can certainly benefit from these tips. Interestingly, most of the money from all of these systems was made on the long side. The best short trade in the one-year system started in 2000 and ended in 2003.
The stock investor could simply move to cash when the one-year strategy signals a sell or reduce long exposure by some measure, such as switching to defensive positions. The commodity trader or the more aggressive stock investor could put on short positions in the S&P 500 futures market.
It must be pointed out that this was a rudimentary treatment of the indicators. A trader possibly could improve his performance by adding other filters or trailing stops or profit targets. None of this was done for this analysis to keep the comparisons between the systems free of non-strategy-based interference.
Short-term traders also can benefit from this knowledge. A day-trader should test a 30-day indicator — today’s price vs. the price 30 calendar days prior. A five-minute trader could look at the current bar compared to the same bar on the prior day — or one 60 minutes or 120 minutes earlier. While the actual strategy may not necessarily hold up on these untested time frames, what most likely will is the tenet that success doesn’t have to happen overnight.
Arthur M. Field has a Ph.D. in management science from Clemson and a J.D. degree from Rutgers. He is a former commodity broker and was co-director of Fidelity’s Pacific Fund and an in-house commodity fund. He wrote “Mastering the Business Cycle and the Markets.”
