AI-powered trading hasn’t yet reached the “iPhone moment,” when everyone is carrying an algorithmic, reinforcement learning portfolio manager in their pocket, but experts say something like that is coming.
In fact, the power of AI comes into its own when faced with the dynamic, hostile arena of trading markets. Unlike an AI agent learning to accurately recognize traffic signals by the endless circuits of self-driving cars, no amount of data and modeling will ever be able to predict the future.
This makes refining AI trading models a complex, demanding process. The measure of success has generally been an assessment of profits and losses (P&L). But advances in the way algorithms are optimized are giving rise to agents that continuously learn to balance risk and reward when faced with multiple market conditions.
Allowing risk-adjusted metrics like the Sharpe ratio to inform the learning process multiplies the sophistication of testing, said Michael Cena, chief marketing officer at Recall Labs, a firm that has run 20 or more AI trading regions, where a community submits AI trading agents, and those agents compete over a period of four or five days.
“When it comes to scanning the market for alpha, the next generation of builders are exploring algo customization and specialization while taking into account user preferences,” Sena said in an interview. “Not just having a raw P&L but being optimized for a particular ratio is kind of the way leading financial institutions work in traditional markets. So, looking at things like this, what is your maximum drawdown, how much was your value at risk to make this P&L?”
Taking a step back, a recent trading competition on the decentralized exchange Hyperliquid, which included several large language models (LLMs), such as GPT-5, DeepSeq, and Gemini Pro, kind of set the baseline for where AI is in the trading world. All these LLMs were given the same signals and executed autonomously, taking decisions. But according to Cena, they weren’t that good, barely outperforming the market.
“We took the AI models used in the Hyperliquid competition and we let people submit their trading agents that they created to compete against those models. We wanted to see if trading agents with additional expertise were better than the basic models,” Sena said.
The customized models took the top three spots in the recall competition. “Some of the models were unprofitable and poorly performing, but it became clear that specialized trading agents that take these models and apply additional logic and heuristics and data sources and things on top are outperforming the base AI,” he said.
The democratization of AI-based trading raises interesting questions about whether there will be any alpha left to cover if everyone is using the same level of sophistication machine-learning technology.
“If everyone is using the same agent and that agent is applying the same strategy to everyone, will it collapse in on itself?” The army said. “Does the alpha it’s detecting go away because it’s trying to execute it at scale for everyone else?”
That’s why those best positioned to ultimately benefit from the benefits of AI trading are those who have the resources to invest in the development of custom tools, Sena said. Like traditional finance, he said, the highest quality instruments that generate the most alpha are typically not public.
“People want to keep these devices as private as possible because they want to protect that alpha,” Sena said. “They paid a lot for it. You saw hedge funds buying up data sets. You saw it with proprietary algos developed by family offices.
“I think the magical sweet spot will be where there is a product that is a portfolio manager but the user still has some say in their strategy. They can say, ‘This is how I like to trade and here are my parameters, let’s implement something similar, but make it better.'”