Case Study Commodities Trading with Market Sentiment

In this case study, we explore how using sentiment data in an investment focused on commodities can significantly outperform a passive strategy. We conducted a backtest on the Gold and Oil WTI, comparing the performance of an active, simple sentiment-powered strategy against the standard Buy-and-Hold.

Case of Gold with GLD ETF

How it is calculated:

The sentiment data was extracted from the Commodities Navigator, an application developed by Alpha Data Analytics that analyzes insights from over 50,000 global sources in 72 languages using artificial intelligence (AI), and GLD ETF daily (close) prices.

Sentiment data has been used for daily portfolio re-balance based on whether market sentiment was positive or negative. Specifically, when week-over-week sentiment outpaced the price, the strategy was to buy and hold until sentiment remained higher than the price. Conversely, when the price exceeded sentiment, the action was to sell existent long and take an equivalent short position. For comparability, both sentiment and price were normalized using their rolling z-scores and rolling to ensure no look-forward biases (leakages) were introduced.

Friction costs and transaction fees are excluded as they are negligible, given the narrow spreads of a highly liquid instrument like a GLD ETF and the infrequent rebalancing, which occurs on average every four days. The period covered the previous year, from February 22, 2024, to the date of writing, February 23, 2025.

Results for Gold:

One-third Better Returns: Sentiment-driven strategy achieved returns one-third better compared to the GLD Buy-and-Hold over the same period. Annual return 37.3% vs. 47%, an excess of +9.64%.

Less Risk: The risk of sentiment approach measured by volatility was 24% lower than the Buy-and-Hold, and 15% lower measured by drawdowns, making it a safer investment option than traditional Buy-and-Hold. Annual Sharpe Ratio 2.02 vs. 3.21 also in favor of sentiment data from Commodities Navigator.

Case of Oil with WTI futures

In the second example, the same strategy based on sentiment and data from Commodities Navigator is used to backtest returns on Oil WTI continuous future contract.

Results for Oil:

Much Better Returns: Sentiment-driven strategy achieved three times better returns compared to the Oil WTI Buy-and-Hold over the same period. Annual return -9% vs. +75.6%, an excess of +84.6%.

Less Risk: The risk of sentiment approach measured by volatility was comparable to Buy-and-Hold, and 31% lower measured by drawdowns, making it a safer investment option than traditional Buy-and-Hold. Annual Sharpe Ratio -0.29 vs. 2.31, also clearly in favor of Commodities Navigator.

Conclusion:

Although gold and oil are both technically commodities, they exhibit a strong inverse correlation of -0.71. Despite this, applying the same strategy over the same period resulted in significant outperformance for both compared to a passive allocation. Sentiment data demonstrated even higher returns, particularly for gold, which experienced one of its best years in history during the case study period, while also reducing risk across all metrics.

The results demonstrate the value sentiment analysis brings to improve investment returns and manage risks. By including market sentiment exported from Commodities Navigator into asset allocation investors can enjoy an advantage compared to conservative approaches. It works and the reasoning behind the approach with sentiment is straightforward: by aggregating all available information and utilizing automated news analysis on a vast dataset, you can assess market expectations and strategically position yourself. This allows you to capitalize on opportunities by buying during oversold periods and selling once positive news is already reflected in the price.

For any questions or feedback, feel free to reach out to us at info@adalytica.com.

Disclaimer: This case study is for informational purposes only and does not constitute investment advice. Past performance is not a guarantee of future returns. Presented backtesting results are the best results on a sample of 3 experiments, other experiment results were comparable.