← Research logMachine Learning
PARKED · unstable
#

Clustering feature vectors (vol, trend, volume, dispersion) into $k$ market states lets each strategy run only in its favourable cluster.

$$ \min_{C}\sum_{i=1}^{k}\sum_{\mathbf{x}\in C_i}\|\mathbf{x}-\mu_i\|^2 $$

K-means / GMM on standardized regime features; map historical strategy performance per cluster.

PARKED
Clusters are interpretable but unstable across refits and sensitive to $k$ and feature scaling; cluster identity drifts, so "trade only in cluster 2" is not robust out-of-sample. Parked pending a stability-constrained clustering. Cousin of the killed HMM (N-032).
Unsupervised regimes are seductive and unstable. Before conditioning capital on a cluster label, prove the label means the same thing next month.

We publish the failures too.

This is one of 100+ documented hypotheses. Browse the full lab notebook, or see the strategies that survived into production.