Knowledge center · Validation & statistics
Factor Correlation
How closely two factors move together in rank, across the same walk-forward samples.
Formula
Correlation(A, B) = Spearman correlation of factor A and factor B scores over shared dates
Interpretation
Two highly correlated factors are not really adding two independent opinions to the composite score — they're one opinion counted twice, which understates true uncertainty.
What good looks like
Most pairs below about 0.5 in absolute value — the factor set is contributing largely independent information.
What bad looks like
A pair above 0.8 in absolute value signals near-duplicate factors; the composite's effective number of independent signals is lower than its factor count suggests.
Typical range
-0.3 to +0.5 for most factor pairs in a well-designed multi-factor model; momentum sub-factors (e.g. different lookback windows) are the most likely to run higher.
How OmniSignal uses it
Directly informs whether the "factor count" backing a score is real diversification or partly redundant — relevant context for reading the factor-contribution breakdown on any research report.
References
Standard pairwise correlation analysis in factor-model construction
Related concepts