All Posts Commodity Trading Week Americas 2024 CTWA24 Digitalisation & Technology

Bringing Order to Crude Chaos: The benefits and challenges of AI/ML applications in crude oil

July 5, 2024

PLEASE COMPLETE THIS FORM TO ACCESS THE content:

Benefits of AI/ML applications in crude oil
Challenges faced in implementing these technologies
Case studies of successful implementations

Amir Azar, Partner, Quantum Horizon Capital

[audio-widget]

Transcript

At this point, I'd love to bring up Doctor Amir Azar, who will be giving a talk on crude markets. What I find really interesting, and we had a great conversation in preparation for this, is how he's using artificial intelligence to analyze markets. I think this is applicable to many different products. So let's have a round of applause for Doctor Azar.

Thank you. Is the mic on? Nice. Thank you, Howard, for the opportunity to be here. The title we have chosen here is “The Crude Oil Chaos,” which I think is an understatement. The oil market often acts contrary to expectations. Today, we will focus on the elements playing in this market, identifying which are quantifiable and how AI/ML can help us navigate this chaos.

Starting with the supply side, let's discuss OPEC production. The cartel's policies, whether to increase or decrease production, may or may not impact oil prices and are highly political. Additionally, we have Russia as part of OPEC Plus, adding complexity.

US rig activities also have a direct impact on crude oil. High prices can create higher rig activities, which can then put downward pressure on oil prices. Another key factor is the strategic petroleum reserves. Though intended for emergencies, recent government decisions to release reserves have interfered with market dynamics.

Geopolitics also plays a significant role. The relationships between OPEC Plus and Russia, the US and China, and crises in the Middle East or the war between Russia and Ukraine add layers of risk. The Strait of Hormuz, a 20-mile waterway through which 20% of oil passes, is a high-risk location given the political conflicts in the Middle East.

On the demand side, global GDP growth impacts crude oil demand and prices. The US dollar's appreciation or depreciation also affects oil prices. China's demand, although somewhat opaque, can be gauged through industrial activities. Renewable energy is another player that will increasingly impact the oil market.

In the US, the marginal barrel of oil, primarily from shale oil, plays a crucial role. The cost of capital significantly influences shale oil activities. Low interest rates during the shale revolution led to a boom in shale production. Elements like the high yield market and investment grade market also affect the cost of capital and subsequently oil production.

To summarize, numerous factors affect crude oil fundamentals and prices, some quantifiable and some not. The question is whether AI/ML can help us position ourselves better. Just to give an example, gasoline prices fluctuate with economic cycles and refining margins. Identifying patterns visually can be challenging, but AI/ML can help aggregate and analyze the impact of these factors.

One way to approach this is by integrating data and using cloud-based systems for flexible computation. Utilizing publicly available data and building proprietary datasets is crucial. Visualization tools can help identify patterns and trends.

Dynamic positioning and limited human discretion are key to managing trade and risk. Systems should adjust to market volatility and minimize human biases. Visualization tools like dashboards can help monitor market conditions and the impact of various elements on prices.

Continuous introduction of new data is essential for AI/ML models to remain effective. Models should be constantly recalibrated and reevaluated to avoid biases. Learning from failed efforts is also crucial. For instance, on February 16, 2024, unexpected market behavior highlighted the importance of accounting for aggregated small details and political decisions.

Happy to answer any questions.

Q: As you train your model, do you exclude outliers and extreme events like COVID-19, which are few and far between?

A: Yes, we do exclude extreme events like COVID-19 from our training because they introduce irregularities that are difficult to model. Excluding such events helps maintain the accuracy and reliability of the model.

Q: In a situation where strategic petroleum reserves are released and sanctions are imposed on major producers, how does your model capture the market reaction?

A: The release of strategic petroleum reserves and sanctions are often politically motivated rather than economically driven. The model can identify elements that positively or negatively impact prices, but predicting such politically driven decisions remains challenging.

Q: How do you manage the training of your model for future black swan events and plan for their impact?

A: Managing for black swan events involves conventional risk management strategies. Increasing risk management costs through options can provide protection during volatile times. It's essential to combine ML-driven and traditional risk management approaches to mitigate unexpected events.

Q: Does your model account for CTA-like activity?

A: No, our model does not currently account for CTA-like activity.

And with that, it's time to break for lunch. Let's give Doctor Azar a round of applause. Thank you for a very insightful and informative presentation.

Written by: Commodities People

AI Commodity Trading Week Americas Crude Oil CTWA CTWA24 ML Technology Applications