In a Sept. 15, 2026 release, SkySail Strategies, a quantitative investment firm focused on applied quantitative research and algorithm development, said its proprietary AI inference model, Rubix RPT, and related forecasting system, Rubix VM, produce market-risk forecasts with 20% to 40% less error than the institutional models that have been used since 1993. The company said the systems are designed to estimate how far a market may move, a figure it described as central to investment risk decisions.
In a study focused on the most volatile days, SkySail said Rubix VM predicted the size of market moves with 98% accuracy. Realized-volatility models used by quantitative hedge funds were 73% accurate, while GARCH models, which the company said have served as Wall Street’s risk-limit standard since 1993, were 49% accurate. SkySail characterized Rubix as roughly twice as accurate as GARCH on those days, with one-twentieth the error.
Brian Devens, managing director of SkySail and a systems engineer specializing in probabilistic and stochastic algorithms, said the legacy risk framework has persisted because it has been difficult to improve. He said the older models trace to research that received the Nobel Prize in Economics in 2003 and that improving on a standard that had held for three decades was difficult but worthwhile. Devens added that the accuracy of the risk number improves every downstream risk decision.
Devens said SkySail is deploying Rubix across more markets and has seen traction from family offices and funds. The company said the model can be applied to any market with a price history.
