⛏️ GDX Elastic Net — Factor Decomposition Dashboard

GDX weekly returns explained by 9 factors | 5 Years | 262 observations

Model

GDX = -0.144 + +1.527×Gold + +0.116×SPY + -0.011×TLT + +0.025×DBC + -0.036×TIP + +0.215×HYG + -0.706×DXY + +0.484×EEM + -0.165×FXI + ε
FactorRaw βStd βOLS βImportance
Gold +1.5268 +3.4739 +1.5337
3.47
SPY +0.1161 +0.2571 +0.1121
0.26
TLT -0.0115 -0.0232 -0.0117
0.02
DBC +0.0251 +0.0675 +0.0242
0.07
TIP -0.0362 -0.0293 -0.0436
0.03
HYG +0.2150 +0.2121 +0.2186
0.21
DXY -0.7060 -0.6466 -0.6989
0.65
EEM +0.4836 +1.1701 +0.4968
1.17
FXI -0.1647 -0.6595 -0.1715
0.66
0.740
α (weekly)
-0.144%
α (annualized)
-7.5%
Best λ
0.001
Best α
0.9
Factors kept
9/9

Raw Betas — Elastic Net vs OLS

Standardized Betas — Factor Importance

GDX vs Gold — Normalized Performance (base 100)

Weekly Residuals — What the model doesn't explain

Weekly
Cumulative
Rolling 12wk
Green = miners outperformed model. Red = underperformed. Cumulative downtrend = persistent alpha leak.

💡 Interpretation

Gold (β=1.53): Dominant driver by far. 1 standard deviation move in gold → 3.5σ move in GDX. Miners are levered gold.

EEM (β=0.48): Significant EM equity exposure. Miners are global-cyclical, not just gold.

DXY (β=-0.71): Strong negative dollar beta. Weak dollar = strong miners. This is partially collinear with gold but adds explanatory power independently.

FXI (β=-0.16): Negative China exposure after controlling for EM. Miners benefit from China weakness — possibly via capital rotation from China into real assets.

TLT, TIP, DBC: Near zero. Rate duration and broad commodities add nothing once you have gold + equity + FX factors. Don't hedge GDX with bonds.

Elastic Net (α=0.9, λ=0.001) with 80/20 train/test CV | Source: Yahoo Finance | Marshall 🔶