Why is the use of algo trading growing in power and gas markets?

There are several underlying reasons for the growth of algo trading, set out below:

• Growth of spot and short-term power and gas exchanges: Algo trading is most easily implemented in liquid markets where credit issues are of limited concern or relevance. Spot and other very short-term trading in power and gas markets is increasingly undertaken on exchanges bringing together market liquidity. These exchanges are much more appropriate venues for algo trading than less liquid, over-the-counter (OTC) markets.
• The growth of renewables and distributed generation: The increasing reliance on renewable energy sources, especially generation connected to the distribution grid, demands response (e.g. large battery units), and the intermittent nature of wind and solar has caused market predictions to require the analysis of a great deal more market data now than in the past.
• Short-term balancing: Typically, market participants are now required to balance their power trading positions in periods of 30 minutes or less, in the context of more complicated grids and in circumstances where prices may be highly volatile. The use of algos can help market participants manage their increasingly time sensitive trading needs.
• Market integration across countries and regions: Power and gas trading is increasingly integrated across countries and regions. The growing number of interconnectors and the rise of market coupling has caused power markets to set the pace in this. However, integration is also occurring in gas markets through interconnectors, increased liquefied natural gas trading (which links disparate markets) and the international use of a small number of benchmarks, such as TTF. These changes again drive the need to analyse vast amounts of data from different markets.

Uses of algos

The increasing complexity of inputs into price formation and the need to execute trades more quickly both favour the use of algo trading.

Specific uses and benefits include:

• Price forecasting and optimisation: Algos can quickly process large quantities of real-time data to provide predictions on future price movements. These predictions can then be used to optimise trading strategies.
• Market monitoring: Algos can react instantly to market signals and execute trades at optimal moments to maximise profitability.
• Arbitrage opportunities through cross-product and cross-market trading: As algos are not focused solely on one market or product, arbitrage opportunities can be detected and exploited in the power and gas markets. For example, algos may be developed to detect price discrepancies between the day-ahead and intraday markets for electricity, and execute trades to profit from this difference. The scope for this is likely to grow as the products within power and gas markets expand and market participants increase cross-border trading in new jurisdictions.
• Risk and exposure management through implementing hedging strategies in real time: Due to the dynamic nature of algos, there may be scope for risk management to develop through machine learning and other artificial intelligence (AI) tools.
• Increasing market knowledge can refine algos: As algos can collect, process and analyse large quantities of real-time data, trading participants may gain better insights into the market and current trends. With this knowledge, algos themselves can benefit from this real-time information, which may increase the accuracy of algorithmic output as they are developed and refined.

Risks arising from algo trading

Although algo trading in the power and gas markets offers a range of benefits for participants, there are significant risks and challenges. These include:

• The potential for disconnects between fundamental market information and algo-driven behaviour: In other words, situations where market performance is driven by how algos react to market developments rather than fundamental market information itself.
• Potential for increased volatility: The rapid response of algos to market signals can amplify existing market movements, thereby increasing market volatility. However, well designed algorithms usually incorporate safety measures to prevent excessive volatility, thereby helping them serve as a useful tool in volatile markets.
• Market transparency and explicability: The increased price movements in markets that may be occasioned by algorithmic trading can make price determination more complicated, especially for manual traders. Algorithms may enhance elements of post-trade transparency documenting individual trading decisions but their complexity – especially where AI or machine learning is involved – may hinder the ability of organisations to explain their trading decisions.
• Algos and market manipulation: While it is argued that by increasing liquidity algo trading can make it more difficult to manipulate markets, poorly designed algos may cause trading that is manipulative or at least disorderly. Although this may be inadvertent, it does not prevent such behaviour falling foul of EU and UK prohibitions since there is no requirement for a manipulative intent.
• Algos using AI: Algorithms may benefit from AI tools to develop machine-learning capabilities. The integration of AI into algo trading carries significant benefits including heightened responsiveness to market dynamics, better adaptability to more complex trading environments and potential ability to tailor trading decisions to individual traders. However, it has the potential to increase the risks highlighted above. In addition, AI may introduce new risks. For example, there is a particular risk that AI or machine learning algorithms may take less reliable decisions in extreme market conditions. Such conditions are rare and may not be included in the dataset on which the algorithm is trained.

Regulatory responses to algo trading: The EU and UK

MiFID II

Algo trading first developed in financial markets and the first regulatory responses to the activity occurred in the regulations governing those markets, especially MiFID II. MiFID II² is a comprehensive regulatory framework for investment services and activities when performed in relation to financial instruments.

Article 17 of MiFID II sets requirements on investment firms engaging in ‘algorithmic trading’³ to have in place effective systems and risk controls suitable to their business and are subject to appropriate trading thresholds and limits. In addition, firms engaging in algorithmic trading must notify their national competent authority (e.g., the FCA in the UK) and provide detailed information about the trading strategies and the types of instruments traded. Finally, MiFID II requires that a firm’s trading systems are resilient and have sufficient capacity in accordance with chapter II of the EU’s regulation on digital operational resilience for the financial sector (DORA). The requirements of Article 17 are supported by more detailed requirements set out in regulatory technical standards (RTS) on the requirements of investment firms engaged in algorithmic trading (2017589/EU) (RTS 6).

RTS 6 sets detailed requirements relating to the governance and control of algorithmic trading, the testing and deployment of trading algorithms, systems and strategies, as well as their management, monitoring and control after they have been deployed.

MiFID II and RTS 6 will be directly relevant to trading in power and gas futures but not to short-term physical trading. That said, as the ACM Study notes, many physical market participants have established compliance frameworks aligned with these standards.

Furthermore, those engaging in ‘high-frequency algorithmic trading’ (HFT) in financial instruments cannot rely on the exemptions from MiFID II that may be relied on by commodity or energy traders (e.g., the ancillary activities exemption). MiFID II defines HFT as an algorithmic trading technique characterised by:

a) infrastructure intended to minimise network and types of latencies, which must include at least one of the following facilities for algo order entry: co-location, proximity hosting or ‘high-speed direct electronic access’;

b) determination by the system (without human intervention) of order initiation, generation, routing or execution; and

c) high message intraday rates (for orders, quotes or cancellations).

REMIT and REMIT II

The EU’s Regulation on Energy Market Integrity and Transparency⁴ (REMIT) sets requirements relating to the disclosure of inside information, prohibition of insider trading and market manipulation, and the reporting of orders and transactions in gas and power markets. In its original 2011 guise, REMIT did not set any specific requirements relating to algorithmic trading.

Recently, REMIT was extensively amended by REMIT II⁵, which introduces for the first time requirements relating to algorithmic trading that are specific to power and gas markets, among other things.

Article 5a of REMIT II sets very similar requirements for market participants as apply to investment firms under Article 17 of MiFID II. However, there is no mandate under REMIT II for delegated legislation to be issued to supplement Article 5a in the same way that RTS 6 supplements Article 17. As the Article 5a standards are drafted at a high level, there is the potential for this to introduce some uncertainty into the concrete steps that must be taken to meet these standards. While this may be addressed by guidance from ACER, we can expect that market participants will use RTS 6 as a benchmark for designing their systems and controls relating to algo trading.

The AI Act

The EU’s AI Act⁶ seeks to regulate the use of AI across different uses, including its application in algorithmic trading within power and gas markets. The provisions most relevant to algo trading are summarised below:

• The AI Act classifies AI systems based on the level of risk they pose, with higher regulatory scrutiny applied to ‘high-risk’ systems. If the AI algorithms used in power and gas markets are deemed high risk, more stringent requirements will be applied.
• More stringent requirements include having comprehensible information on how the high-risk AI system was developed and its performance⁷. In the context of algo trading, this is likely to involve detailed record keeping, including how the AI makes trading decisions, what data sets were considered, and whether there was any human oversight involved. Although this appears to be broadly in line with MiFID II and REMIT II requirements, there could be additional documentation obligations.
• The AI Act focuses on the importance for high-risk systems to be designed in a manner that enables deployers to understand how it works, including instructions for use. There would need to be relevant human oversight measures and, in practice, may require human traders to regularly review the AI’s trading strategies and decisions.⁸
• Finally, the AI Act requires market users to be able to justify and explain trading decisions conducted by AI algos. This is directly relevant to the need for a level of human oversight and accountability.

United States – Regulation

The United States does not have the same type of prescriptive rules related to the types of algorithmic trading typically used in the power and gas markets. For example, neither the Commodity Futures Trading Commission (CFTC) or the Federal Energy Regulatory Commission (FERC) have licensing requirements related to the use of algorithmic trading strategies in their markets.

The CFTC does impose requirements on futures exchanges related to the prevention of market anomalies and disruptions caused by algorithmic trading,⁹ but this is generally seen as codifying risk control requirements that futures exchanges already had in place. Specifically, the CFTC requires U.S. futures exchanges to: (1) implement rules designed to prevent, detect, and mitigate market disruptions and system anomalies associated with electronic trading; (2) implement pre-trade risk controls for all electronic orders; and (3) promptly notify the CFTC of any significant market disruptions on their electronic trading platforms.¹⁰

However, market participants using such strategies are responsible for the trading activity conducted by such programs, and many CFTC and FERC enforcement actions relate to automated trading programs that violate regulatory requirements (e.g., spoofing). Additionally, respondents in such enforcement actions (and disciplinary actions from futures exchanges) are often charged with a failure to supervise.

Separately, certain U.S. futures exchanges have certain technical requirements related to algorithmic trading, which are important to understand. For example, certain futures exchanges require automated trades to be reported with specific identifiers, or tags, indicating that they were executed via an algorithm rather than manually. Other rules may require traders executing manual and automated trading strategies to use different tags identifying the trader and whether the trade was executed manually or via an algorithm.

Final Thoughts

The use of algorithmic trading in power and gas markets is likely to continue to increase, as it can bring significant benefits to market participants. However, it can also carry significant risks. Regulators have responded to these issues in a variety of ways, and we expect further regulatory and legislative change, especially as AI and machine learning are deployed in connection with algorithms.

1. “Algorithmic Trading in Wholesale Energy Markets: Key findings of an exploratory market study by the ACM”
2. 2014/65/EU. As an EU Directive, MiFID II was implemented in the UK and there have been very few changes to the substance of the relevant requirements since the UK’s withdrawal from the EU.
3. Article 4(1)(39) of MiFID II defines ‘algorithmic trading’ as “trading in financial instruments where a computer algorithm automatically determines individual parameters of orders such as whether to initiate the order, the timing, price or quantity of the order or how to manage the order after its submission, with limited or no human intervention, and does not include any system that is only used for the purpose of routing orders to one or more trading venues or for the processing of orders involving no determination of any trading parameters or for the confirmation of orders or the post-trade processing of executed transactions.”
4. Regulation (EU) No 1227/2011.
5. Regulation (EU) 2024/1106.
6. Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence.
7. Para 71, Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence.
8. Paras 72 and 73, Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence.
9. See Electronic Trading Risk Principles, 86 Fed. Reg. 2048 (Jan. 11, 2021).
10. 17 C.F.R. sections 38.251(e)-(g).

In-depth 2024-192