Seasonality is a well-known phenomenon in the commodity markets, with certain sectors exhibiting predictable patterns of performance during specific times of the year. These patterns often attract investors who aim to capitalize on anticipated price movements, creating a self-reinforcing cycle. But what if we could stay one step ahead of the crowd? By front-running these seasonal trends—buying sectors with expected positive performance (or shorting those with negative seasonality) before their favorable months begin—you can potentially gain a significant edge over traditional seasonality-based strategies. In this blog post, we explore how to construct and backtest a systematic strategy using commodity sector ETFs to exploit this seasonal front-running effect.
Seasonality affects both the supply and demand dynamics of commodities. For instance, agricultural commodities like corn and soybeans exhibit price fluctuations based on planting and harvest cycles, while seasonal consumption patterns influence energy commodities such as natural gas. Understanding these seasonal trends gives traders a strategic edge, allowing them to anticipate price movements and make informed trading decisions.
Three example sources examine the nuances of seasonality in commodities, employing quantitative methodologies and econometric analyses to elucidate the phenomenon. Utilizing these studies highlights the intricate dynamics and cyclical behaviors observed in financial markets.
Our following research, which will be presented, further deepens the existing body of knowledge on the commodity seasonality topic and provides practical implications in the form of a sound and easily executable strategy.
Based on the utilized literature review, we can make an educated guess that seasonality will significantly impact the cross-section of commodity sectors. We will introduce our investment universe selections through the data, move on to formulating our initial hypothesis, and follow it with rigorous testing in the next chapter.
We utilize the same asset universe as in the study referenced in our How to Improve Commodity Momentum Using Intra-Market Correlation. ETFs are simple, have long historical data, and represent sectors, thus reducing idiosyncratic risk. This approach minimizes the impact of individual commodities and emphasizes the overall effect of the entire sector, potentially simplifying the identification of trends and seasonality. Nevertheless, we hypothesize that this seasonality effect will apply to individual commodities/ETFs. Data can be easily obtained in daily granularity from Yahoo Finance.
| DBA | Invesco DB Agriculture Fund |
| DBB | Invesco DB Base Metals Fund |
| DBE | Invesco DB Energy Fund |
| DBP | Invesco DB Precious Metals Fund |
The complete investment universe consists of 4 ETFs, data analysis starts in January 2007 and ends in June 2024.
Our initial hypothesis was that:
The performance in month X could be predicted by the performance in month X-12.
For instance, we could predict February’s performance based on the performance of the previous February.
At the end of January, we seek a signal indicating whether to hold the commodity sector in February. We evaluate February’s performance from the previous year and examine its percentile rank among the monthly performances over the last 12 months.
Subsequently, we calculate this for all four commodity sectors. And we then
Let’s give an example. At the end of January, we examine the percentile rank for each ETF (DBA, DBB, DBE, DBP) among the monthly performances over the last 12 months. DBA’s performance over the previous year’s February is 25% percentile of all 12 months. DBB’s performance over the previous year’s February is 75% percentile of all 12 months. DBE’s performance is in the 83% percentile, and DBP’s is in the 83% percentile, too. Therefore, at the end of January, we will go long 3 ETFs (DBB, DBE, and DBP), each with 25% weight, and we will short one ETF (DBA), with -25% weight. As you can see, a total net position for the strategy can be between -100% (we are short all 4 ETFs) to +100% (we are long all 4 ETFs).
An advantage of the construction of the seasonality strategy in this fashion (net ETF position varies between -100% and +100%) is that it better captures seasonal patterns of the commodity sectors. It is easily possible that some months of the year have negative seasonality for the majority (or all) commodity sectors. On the other hand, it is quite possible that some months of the year have positive seasonality for all of the commodity sectors. This time-series seasonality definition judges each commodity sector on its individual basis, not on a cross-sectional basis against other commodity sectors. Of course, it’s also possible to build the cross-sectional seasonality strategy (ETFs in the investment universe are judged against their peers), and we may revisit this cross-sectional seasonality strategy in later articles.
And here we have the outcome of such simple, naive seasonality strategy in comparison to the benchmark (equally weighted portfolio of 4 ETFs).
| CAR p.a. | Volatility p.a. | Sharpe Ratio | Max DD | CAR / (max DD) * | |
|---|---|---|---|---|---|
| Model Portfolio 1 |
-4.54% | 11.26% | -0.40 | -60.26% | -0.08 |
* also called the Calmar ratio or Return Over Maximum Drawdown (RoMaD)
It’s quite a disapointment right? The strategy has a significant negative drift over whole backtested period. It looks like there is even a reversal in the seasonality – previous year’s February negatively predict performance of the current February. But why it’s so. Our hypothesis is simple – sophisticated market participants know about the strong seasonal tendencies in commodities and are front-running seasonal signals.
Our hypothesis posits that instead of determining the performance in February compared to the entire previous year at the end of January and using that as the basis to buy in February, it is about front-running the signal. Hence, at the end of January, we should be concerned with the anticipated performance in March. This would enable us to purchase the corresponding sector at the end of January and hold it until the end of February, thereby front-running the seasonality by one month, as sophisticated players in the market would do.
So, let’s adjust the seasonality parameter in our strategy in a way that we will try to front-run seasonality signal from the initial version of the strategy – at the end of January, we will buy/sell commodity sectors based on their March performance (compared to the last 12 months of data).
Our alternative hypothesis is:
The performance in month X could be predicted by the performance in month X-11.
Let’s look at the picture below. It’s quite different if we compare it to the previous version. The newly developed trading strategy has improved quantitative metrics, confirming our alternative hypothesis.
We can clearly see that there must be a front-running effect within commodity sector seasonality patterns, particularly in commodity sector ETFs. A trading strategy that trades ETFs with an X-12 seasonality signal has a significantly negative performance, while a strategy with an X-11 seasonality signal has a significantly positive performance. It would not be surprising if similar patterns were also observable in individual commodities.
Performance metrics for 2nd Model, comparing It to 1st Model
| CAR p.a. | Volatility p.a. | Sharpe Ratio | Max DD | CAR / (max DD) | |
|---|---|---|---|---|---|
| Model 1 |
-4.54% | 11.26% | -0.40 | -60.26% | -0.08 |
| Model 2 |
6.71% | 12.10% | 0.55 | -20.43% | 0.33 |
As you can see, there is a significant performance increase and portfolio has better risk metrics overall.
What can we say at the end? It may be a good idea to include a seasonality patterns in our trading decisions. They may not be the strongest predictors that are available to us, but it’s not wise to neglect them. On the other hand, financial markets are adaptive, so the naive seasonal trading strategies may not perform the best.
Our second trading strategy, incorporating proactive seasonality time-shift, has significantly enhanced performance metrics and overall effectiveness. By shifting our seasonality parameter to anticipate market movements one month in advance, we have aligned our strategy with the behaviors of the most sophisticated market participants. This adjustment has resulted in markedly improved outcomes, and these results underscore the importance of continually evolving trading strategies to stay ahead in increasingly efficient markets.
Our recent examination has highlighted the influence of intra-market correlation on commodity momentum. Drawing from the insights of our existing Quantpedia article, How to Improve Commodity Momentum Using Intra-Market Correlation, we extended our hypothesis that correlation structures can significantly influence also our new seasonal strategy efficacy.
Findings Recapitulation:
Correlation Hypothesis Confirmation:
Our analysis confirmed that the correlation between short-term and long-term returns can indicate success in seasonality strategy.
Graphical Analysis: Graph showcasing the seasonality strategy relationship with newly defined predictors
| CAR p.a. | Volatility p.a. | Sharpe Ratio | Max DD | CAR / (max DD) | |
|---|---|---|---|---|---|
| Model 2 + corr filter |
7.62% | 9.43% | 0.81 | -11.98% | 0.64 |
| Model 2 |
6.71% | 12.10% | 0.55 | -20.43% | 0.33 |
This is quite a surprise. The short- vs. long- correlation as a predictor was used by us in the previous article in which analyzed commodity momentum and we had a suspicion, that it may improve also seasonality strategies. But we didn’t expect it to improve seasonality strategy so much. The main hypothesis as to why the seasonality strategy performs better when the short-term correlation is lower is due to the higher dispersion of the commodity market in this state. When the short-term correlation between commodity sectors is low, it’s evidently easier for the seasonality strategy to pick/distinguish the seasonal signal from the noise than in the state when the correlation among commodity sectors is high (and they may trend all in the same direction).
Author: Cyril Dujava, Quant Analyst, Quantpedia
Are you looking for more strategies to read about? Sign up for our newsletter or visit our Blog or Screener.
Do you want to learn more about Quantpedia Premium service? Check how Quantpedia works, our mission and Premium pricing offer.
Do you want to learn more about Quantpedia Pro service? Check its description, watch videos, review reporting capabilities and visit our pricing offer.
Are you looking for historical data or backtesting platforms? Check our list of Algo Trading Discounts.
Or follow us on:
Facebook Group, Facebook Page, Twitter, Linkedin, Medium or Youtube
Share onLinkedInTwitterFacebookRefer to a friendBe first to know, when we publish new content
