QM011 Hands-on Lab
Strategy Similarity and Decision Diversification
Objective
Compare two synthetic allocation paths along three distinct dimensions: what they hold, when they make a material change, and how far their weights differ. The exercise shows why return correlation alone cannot describe decision similarity.
Run
From the QM011 folder:
python labs/python/qm011_decision_similarity.pyWhat to inspect
The script reports period-by-period weight overlap, action flags for each strategy, timing disagreement, and average half-\(L_1\) weight distance.
- High overlap in one month does not imply that the two strategies change at the same time.
- Timing disagreement depends on the material-change threshold
tau. - Half-\(L_1\) distance measures allocation disagreement, not expected-return superiority.
Experiment
Change one row of A or B, then change tau from 0.2 to a lower or higher value. Identify whether the change affected the what, when, or how-much dimension. Avoid collapsing the diagnostics into one score unless a specific decision rule requires it.
Interpretation boundary
These are descriptive decision-similarity diagnostics on synthetic weights. They do not establish diversification benefits by themselves; portfolio admission still requires evidence on realized interaction and role fit.