Can AI discover the hidden laws of markets?
Financial institutions are sitting on massive reserves of high-dimensional, noisy data. Standard AI models can often predict patterns within this data, but they remain “black boxes”. They are unable to explain why those patterns exist or what rules govern them.
At Sigma AI, we specialise in a different approach: turning under used data into interpretable, actionable mathematics.
We constantly monitor the bleeding edge of AI research to find methods that don’t just forecast, but actually explain complex systems.
Our team took a brand-new method for finding governing equations originally designed for neuroscience – and pointed it at equity order-book data. While it did not out-forecast standard models, it independently rediscovered three fundamental laws of markets, and wrote them down in plain mathematics. The results below showcase how Sigma AI can take raw, unlabelled data and extract the hidden laws governing your most complex systems.
Key points
We adapted DYSCO, a 2026 method for finding governing equations in noisy data (until now tested only on simulated systems), to work with large-cap equities and live order-book data.
It never out-forecast simple baselines, but it did rediscover, on its own, three textbook laws of markets: volatility clustering, the leverage effect, and bursty order flow.
A black-box model might learn the same patterns but can’t tell you what they are.
The approach here writes them down as interpretable equations.
Do you want to unlock the full content?
