AI facilitates Brady’s automated trading
AI is the defining word of our time. Conversations about AI permeate every aspect of the energy business from coding assistance, data analysis, and forecasting to interface development, risk management, back-office operations, decision support, natural language processing, and reporting. But what is truly transformative, and where are people simply jumping on the train out of fear of missing out?
One of the biggest concerns revolves around AI-driven trading decisions. Since trading decisions can lead to significant losses, accountability for such losses must ultimately rest with humans. I recently had a discussion with Chris Regan, the Managing Director of Brady, about AI-powered algorithmic trading trying to identify the exact role of AI in the process and quantify the gains AI usage brings.
Most algorithmic trading solutions focus on automated rule-based intraday trading. They process real-time market information and position data, using rule-based algorithms to make trading decisions. A human trader always retains a kill switch to halt execution. My question was: where does AI fit into this process?
Chris explained that Brady’s algorithmic trading solution isn’t a set of fixed, pre-coded algorithms. Instead, it offers a library of Python-coded components, forming an ecosystem that customers can orchestrate into their own trading strategies. The real innovation, according to Chris, lies in the integration of AI-driven forecasting into certain elements of this library.
Machine learning is already widely used in predictive analytics, so incorporating AI-based predictions into algorithmic trading feels like a natural next step. If an algorithm can anticipate market trends with a certain level of confidence, it can enhance the profitability of automated trading. Here, AI’s role is to learn from historical market data, including seasonal patterns and recognize scenarios in which prices are likely to rise or fall with certain confidence.
Most companies use algorithmic trading to automate efficient position closing. In these cases, the mark-to-market value typically hovers around the average or slightly below, as rule-based algorithms prioritize minimizing losses rather than maximizing profits through aggressive trading. As Chris puts it, rule-based trading results in a “mark-to-average” strategy.
However, integrating predictive mechanisms shifts the focus from efficiency to profitability. According to Brady’s back-testing results, AI-enhanced trading can secure prices up to 5% better than traditional rule-based approaches. That said, this strategy comes with additional risk. The profit-and-loss distribution develops a heavier negative tail, meaning traders must determine how much risk they are willing to accept in pursuit of higher average returns.
Chris brought an example of a company which was primarily seeking to maximize profits through arbitrage between day-ahead and intraday trading, focusing on value protection within intraday markets. Even in such case, it was possible to design an intraday strategy with minimal risk while still achieving slightly better-than-average M2M performance, Chis told me.
Keep in touch and sign up to our Newsletter
