The best months to be long or short

January 29, 2017 09:00 AM
Seasonal tendencies often refer to agricultural markets, but all markets provide seasonal clues.

Seasonality is one of the cornerstone approaches to futures trading. There are undeniable biases that can be exploited for fundamentally obvious reasons. Spring is when corn is planted and fall is when it’s harvested. It would be logical to assume distinct market action peculiar to each given time of the year.

The old bellwether is that agricultural commodities best lend themselves to textbook analysis, but most, if not all, markets also reflect time of year long/short biases. This is a generalization, of course, that will not necessarily reflect a given year. While September and October may be the weakest months of the year overall for stocks, 2013 saw a 126.25 point rise in the S&P 500 during that timeframe. This wouldn’t flummox any entry level statistician — there are tendencies and contrary deviations within them.

These variances are not all that infrequent and that’s one reason seasonal trading tends to be easier in theory than in practice. It’s very hard to stay with something on which you’re risking thousands if not tens of thousands of dollars. Even trades that play out near perfectly according to expectations will likely have wide interim swings that will take out even moderate stops. That’s one downside to the approach’ and there are others.

Still, if there are seasonable biases, why not be aware of them? Any knowledge only adds to the tools in our arsenal. Perhaps one doesn’t abandon an existing methodology in favor of a seasonal system, but rather uses the seasonable bias as a filter. You may weigh your exposure more heavily when the trades match the seasonality, or completely eliminate those moving in the counter direction. 

The easiest way to test for seasonality is to optimize for best monthly buys and sells. A simple TradeStation strategy follows. Its reserve function “month (date)” assigns the January through December months 1 through 12. We set up a monthly chart and insert the following code: 

  • inputs: y(9),z(2),p(0);
  • variables:m(0),x(0),mp(0);
  • m=month(date); {1=January, 2 February, etc. up through 12 December}
  • mp=marketposition;
  • If m=y then buy next bar at market;
  • if m=z then sell short next bar at market;
  • If mp=1 and barssinceentry=p then sell next bar at market;
  • If mp=-1 and barssinceentry=p then buy to cover next bar at market;

We’re optimizing for two variables together in the same study. One is the month (m) in which we enter a buy. This will correspond to the input value of “y.” Actually, we will always be entering one month after the assigned value because the code reads “buy the next bar at market.” Nine represents September — the bar we’re in now so we’ll buy at the opening of the October bar. 

    The same is true for the sell side, our second optimized input, represented by “z.” We’re showing 2, or February, meaning that we’ll be entering at the opening of the March bar.

Included is a third input: “p,” which would allow you to optimize the number of months you would hold the position. For this demonstration, though, we’ll just assume we’re trading from the open to the close of a single month (p=0) and we’ll optimize only for the best long and short entries. This will hold the trials to a manageable number and will also eliminate the issue of trades getting reversed by an opposite signal before they can fully play out.

The formula included in this article reflects the best optimized numbers in the 30-year U.S. Treasury Bonds (see “Season of the long bond,” below). You’d be long throughout August, (y being 7 or July, and you’d buy the next month on the opening). The best short month is October (Z=9 or September—sell short on the first October opportunity). 

“Inside the numbers” (below) is the optimization study showing the 144 possible combinations from best to worst. We’ll obviously discount the 12 times when buys and sells occur in the same month. The tests are from best to worst according to the return on account; net profit divided by the worst drawdown. This is the best stat to optimize as it gives you performance in relative terms — not only what you ultimately would have made, but the pain you would have had to endure at the worst level. The first trade was October 2002. The final, October 2016, trade produced a striking $5,625 profit. Each trade was docked $5 slippage/commission.

The net $51,511 profit is impressive, as is the fact that even without stops, you would have had no worse than an end of day $5,582 drawdown. Still, we have to be mindful of how daunting holding a non-stop strategy over an entire month can be. 

Clearly, we’d want to incorporate some sort of stops, filters or other qualifiers before we could treat this as a bona fide turnkey system. Note also that a monthly study is inevitably going to mean a relatively small number of trades. If you’re going to count on the veracity of a mere 29 trials, you’d be advised to also have some other knowledge at hand such as a plausible theory of what to expect from the yearly cycles. 

Soybeans offer more traditional seasonal tendencies. Traders would all want to play for the summer anomaly — drought, flooding, anything that would ruin the crop and drive prices to career-making heights. Year after year, traders would be disappointed as the growing season played out without incident. 

If anything, you want to be a seller, not buyer as summer gets into its dog days. For the last three years, a short placed on July 1 could have eventually taken $2 out of the market ($10,000 per contract). The worst heat against you would have been less than $1,000.

While summer rarely produces the hoped-for weather scare, spring frequently does. It’s generally not a similar $10,000 opportunity and in fact, you may have to trade very skittishly — hardly a quantifiable recommendation. The point is, there is almost inevitably some sort of scare, albeit a minor one, in the spring while crop progress is more anticipated than actualized.

Holding longs in April, shorts in July produced a $46,257 profit since 2004; $29,585 was from the long side, $16,672 from the short (see “April showers …,” below). As you can see in the yearly breakdown, the bias has since held up nicely.

Interestingly, though, the best return on account results are far afield of April/July. “Beans, offseason,” (below) reveals that your best result for the last 14 years would be to hold longs in December and shorts in September. 

I can’t offer a fundamental justification.

Testing in all five major U.S. stock related futures markets confirms the widely held belief that September-October is a particularly good shorting timeframe. Among the contributing factors: Black Monday, Oct. 19,1987; 9-11 and Black Friday, Oct. 13, 1989. Since 1980, six out of the last seven worst percentage drops in the Dow Jones Industrial Average occurred in either September or October.

Nevertheless, testing reveals that since 1999, as far back as one can test in TradeStation, there is an even better overall month for shorting the S&Ps. That would be January. The best long month immediately precedes December. 

“Season for stocks,” (below) shows an S&P performance summary and optimization result. 
The Russell testing field goes back to January, 2003. As in the S&P, its best aggregate shorting opportunity occurs in January. The December long bias takes second to March, the third best long month in the S&P 500.

Again, seasonal biases almost certainly won’t be all you need for construction of a viable trading system. They do, however, provide stats that could be useful. Stats almost always are.

About the Author