Questrade Financial Group
Detecting Market Regime Shifts to Support Risk-Aware Investing | Student Capstone Project
Questrade Financial Group is one of Canada’s leading online brokerages. Managing over $50 billion in assets, the company faces a common industry challenge—navigating unpredictable market swings.
While financial markets naturally shift between calm and turbulent periods, detecting those changes in real time is difficult. Early detection of market regime changes can give portfolio managers critical foresight that helps them better manage risk and improve investment outcomes.
Questrade Financial Group worked with a group of UBC Master of Data Science Okanagan students to develop a system that provides timely, actionable insights by detecting these shifts weeks before significant declines.
To begin, the students analyzed 23 years of daily return data from SPY, an ETF (Exchange-traded fund) that tracks the S&P 500 and reflects broader market behaviour. Then, they focused their efforts on building a dual-model system to capture both gradual and sudden changes in market conditions.
The system included Markov-switching GARCH (MS-GARCH), a probabilistic time-series model that captures volatility patterns while forecasting future market risk, and Change Point Detection, a method that flags sudden disruptions in market behaviour that are often triggered by events like financial crises or major policy changes.
After preparing the data, both models were validated through statistical testing and backtesting against historical market events to ensure they reliably detected real market regime shifts—and they did.
In a 10-year simulation of a simple investment strategy, the models doubled investment returns compared to standard strategies, delivered steadier gains while lowering overall risk, and helped avoid millions in potential losses during market downturns.
By developing a system that detects early and reliable warnings of market shifts, the students added a valuable layer of foresight to Questrade’s risk management process.