In the first installment of this two-part series, we discussed the intermittent shocks affecting the price of real commodities. We looked at hogs as an example of dramatic price swings and then built a simple model to trade these price shocks using live cattle futures. The model was based upon a three-month moving average of monthly closing price momentum, expressed as a percentage.
Building this model is easy with an Excel spreadsheet and clean historical closing data. After the model is built, data are updated at the end of each month. This is an excellent tool in the part-time trader’s arsenal. It produces long-term signals and is always in the market. For example, if the cattle model signals a sell order in March 2010, the trader can enter a position and hold short until the next buy. Alternatively, other short-term strategies can be oriented in the direction with the long-term signal, providing a certain level of comfort due to duplication of signals from different methods.
As a quick refresher, we know that price is dependent upon supply and demand and external events occur without warning, sending prices soaring or crashing. Some events are significant, causing a severe imbalance in price, which can take longer for supply or demand factors to respond accordingly. It is these shocks, and the responses to them, that we seek to trade.
Analysis reveals that different markets measure and react to shocks in different ways. The basic model for cattle produced 10 completed trades over a 15-year period with 100% accuracy. It did suffer drawdowns in some trades, but each was completed with a profit, averaging in excess of $7,000, after slippage and commissions. Adding a second “trading” contract increased the average net profit to over $10,000.
Now we’ll turn to other commodities and build slightly more complex models. The last article depicted major turning points in hogs since 1994 using high and low prices and produced 423.48 points of swing.
For our new model, we start with the same simple three-month moving average of momentum, our cattle model, but we observe, as with most commodities, that prices tend to explode on the upside more than they crash on the downside. This may result from a speculator short squeeze, a demand panic or a true depletion of supply. Generally, a combination of these generates the extreme upward price increases we’ve seen.
So, we set bands at different levels if needed and assume the sell level may be greater than the buy trigger. Levels of 11.4% and -8.4% are good points for selling and buying, respectively. The model generates nine completed trades for a gross profit of 80.80 points, six wins averaging 14.5 points each and three losses averaging 2.1 points. This may not be the best we can achieve.
Consequently, let us construct a second statistic: one-year price momentum. Compute a percentage momentum statistic of this month’s price close vs. the same month the prior year. For example, in June 2008, hogs closed at 71.77¢. In June 2009, hogs closed at 58.67¢, a drop of 13.10¢, or -18.25%. When we analyze this statistic, we see it is highly dependent upon any prior shock in price, but still generates 135.80 points over five profitable trades. We might still do better.
To soften this shock, we generate a modified statistic. First, compute a three-month moving average of price for the prior year. Next, generate a momentum statistic for this month divided by the average generated 11 months ago — that is, the average of prices 11, 12 and 13 months ago. We call this the “hog model.”
We will set our bands at 49.5% sell side and -34.0% and -49.5% on the buy side. This time, they will be penetration bands. Instead of waiting for a retracement after momentum exceeds the trigger level, simply execute the trade any time the statistic trades beyond the key number.
For example, consider the set: -7.4%, 36.2%, 38.1%, 50.3%, 54.9%, 33.3%, 61.4%, 62% and 21.3%. In those months, assuming we are long, we will sell on the close of the 50.3% month, not wait for the 33.3% retracement. In addition, because the statistic fell below the 49.5% level and then climbed above it again, add another sell position at 61.4%. As a result, our five profitable trades now yield 254.87 points of profit. And, on the buy side, we will go long at a penetration of -34%, and again if in a subsequent month price plunges and the statistic exceeds -49.5%, a phenomenon we don’t expect to see except in unusual circumstances.
As with cattle, add a second contract on the initial reversal trade with a profit target of 17.50¢. If the signal repeats, add a contract, but do not add the profit target contract unless the prior extra contract has already satisfied its target. Doing this raises the gross profit to 361.43 (see “Two contracts are better than one” ).
The model produces nine profitable trades with an average gain of 40.15 points. Assuming cost and slippage of a whopping $500 per trade, that is still over $140,000 profit for the model — roughly $9,300 per year. Even though the drawdowns can be difficult, such as from July 1998 to the low in November 1998 (about $16,500 on the two contracts), the model does successfully catch the peaks and valleys. As of July 24, 2009, the model remains short one contract from 79.07¢ with a closing price of 59.05¢, for an unrealized gain of an additional 20 points.
ENERGY MARKET MAYHEM
History has ample evidence of crude oil shocks, which also tend to be a recession predictor and have a utility beyond a mere trading vehicle. Spikes in oil correlated with nine out of the 10 recessions since World War II and they are a leading indicator for peaks in unemployment.
In October 1988, prices were at $11 per barrel and jumped to $22 in December 1989, just about 8% per month. In 1998, prices were at a low of less than $11 in June. They increased through September 2000 to more than $35 per barrel, 218% in just over two years, an average of 8% monthly. In 2006, oil prices rose from $58 to $77 in roughly seven months, only a 5% increase per month and beneath the level of a historic recession shock. In 2007, oil hit a bottom of $49.90 on Jan. 31, 2007. On Jan. 4, 2008, crude hit $100 per barrel, a monthly increase of 8.3%, which is the excessive run to predict a recession start. Price declined 14% over the next 21 days. Oil topped in July 2008, just less than $148 per barrel, then crashed to $33 per barrel in less than six months and then bounced to $80, a typical oil shock followed by a crash and a rebound.
Again, begin with the simple cattle model, using band triggers of 9.4% sell and -7.5% buy. Starting in January 1984, the basic model produces 11 wins and two losses for a respectable gross gain of 193.39 points in 21 years. A one-year momentum model with a retracement buy trigger at -35% and a sell level of 80% has just six trades, all profitable, for 328.77 points (see “Momentum in crude”).
Admittedly, it takes 23 years to make 328 points, but that is still an average of $14,295 per year. Even allowing $500 slippage per contract, the model realizes roughly $324,000 net profit with little drawdown.
Lumber is another commodity that’s highly correlated with the business cycle. Applying the cattle model to lumber using bands of plus or minus 8.4% produced a gross point gain of 926.10 points from January 1993 through June 2009, roughly $101,000 from nine winning trades. Using the hog model with optimized retracement triggers for shorts of 55% and longs of –32.75% produced four winning trades for 1,431.80 points, or $157,000 with less drawdown. Adding the extra trading contract with a $150 profit target boosted the gross gain to 2,051.10 points, about $225,000 (see “Lumbering profits”). Both long contracts are open with a closing price of $188.00 on July 24, 2009. The profit target is $297 for the trading contract.
Lumber is an excellent predictor of the business cycle. The January 1997 sale coincided with the high GDP reading of 4.5% and the September 2000 low preceded the official end of the recession. Similarly, the 2004 sales coincided with the high gross domestic product reading of 3.6% and the February 2009 low may be an early sign of the end of the 2008 recession.
Soybeans are attractive, with active 5,000 or 1,000 bushel contracts. The basic cattle model produces 2,648 points of gain with five profitable trades from April 1987 to June 2009. Bands are set at 9.7% and -8.4%. The hog model does better using penetration bands. A long trade is signaled when the statistic penetrates -31.5%; a short trade when the statistic penetrates 70.25%. However, unlike the other commodities we have modeled, do not re-enter the trade on the same side if the statistic trades beneath the band and then above it again. Instead, put on a subsequent position only if the statistic penetrates -50% for longs, or 100% for shorts, indicating a massive blowout. Aggressive traders can place a duplicate extra trading contract at the initial signal and exit on a profit target of 100 points. As a result, we gain 3,654 gross points (see “Sweet on beans”). The remaining long bean contract is open and profitable.
It is clear from the above examples that price shocks make excellent reversal points for the market and a model can be constructed to profit from these on a long-term basis. Consider testing other commodities for similar profitability. Although we have examined only monthly data, it is possible similar methodology could be applied to a five-minute chart for the day-trader seeking to take advantage of sudden aberrations in price.
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.”