SVD Market Regime Detection Framework
1 min read
Placeholder
Placeholder one-sentence summary of the SVD regime detection approach.
Regimes detected
4
Backtest Sharpe
1.8
Universe size
120+
Placeholder
Challenge
So I noticed regime shifts were getting buried in correlated noise across assets, and identified that traditional clustering was too slow to react when volatility spiked.
Solution
Built an online SVD pipeline that tracks dominant eigenvectors as regime fingerprints, with thresholds tuned on rolling validation windows.
Placeholder
PythonNumPyPandasPlotly
Result visual