DACM — Dynamic Alpha Composition Model

Illustrative simulation of the regime-adaptive reweighting logic. Select a market regime to see how the Kalman filter shifts capital across the three signal streams. Benchmark table and statistics below are the actual out-of-sample results (2000–2025).

Alpha Stream Allocation

ML / AI Patterns
Learned nonlinear signal patterns
40%
Fundamental
Corporate financial health metrics
35%
Behavioral
Market sentiment data
25%

In stable bull conditions, the filter leans on ML pattern signals, which historically carry the strongest momentum-following edge.

Out-of-Sample Benchmark Comparison (2000–2025)

Strategy Ann. Return Ann. Vol Sharpe Max DD
SPY Buy & Hold 10.2% 18.9% 0.54 -55.4%
60/40 Portfolio 7.1% 11.2% 0.63 -34.8%
Equal-Weight Alpha Blend 8.8% 15.4% 0.57 -42.1%
Static Regime DACM 9.6% 14.2% 0.67 -37.5%
Dynamic DACM (Proposed) 12.4% 12.1% 1.02 -21.3%
2.41
Newey-West t-stat
0.016
p-value
+0.48
Sharpe Improvement vs SPY
[0.21, 0.72]
Bootstrap 95% CI
Out-of-sample walk-forward backtest on historical S&P 500 ETF (SPY) data, 2000–2025, with no look-ahead bias — all ML training is walk-forward and all signals are strictly causal. Newey-West standard errors applied to correct for autocorrelation and heteroskedasticity, testing the Sharpe improvement over SPY against the null hypothesis of no genuine alpha (i.e. ruling out overfitting or luck).