Historical Data

disabled for public users

Waiting for Python runtime...

Last 15 rows

Sentiment

Three features are z-scored over a rolling window of selectable length, then reduced to a single component via PCA. PC1 is the "sentiment" index. Changing the window below recomputes PC1 (and re-fits the regimes on the Results tab).

Select rolling window:

PC1 (Sentiment)

Waiting for Python runtime...

Feature Loadings on PC1

Regimes

Fit a Gaussian Mixture Model on PC1 to classify each day into a regime. States are predicted across all available history regardless of the training window.

Inputs

PC1 available: --. Leave dates blank for full series.

Number of states

Train start

Train end

Return Distribution by Regime (S&P 500)

Annualize returns

Regime Highlights (S&P 500)

Log scale (decompound)