Purpose
Probability estimate for 30-day positive BTC returns using a 10-category weighted scoring system with regime adjustment, Weekend Effect indicator, and lagged predictors.
Performance (Out-of-Sample 2013-2025)
Category Weights (Wild Bootstrap OLS, 40k iterations)
Category
Weight
% Sig
Interpretation
Technical
2.68
100%
★ STRONGEST (t=13.34) - 20/50d momentum, MA200, RSI
Sentiment
1.64
100%
Fear & Greed contrarian (t=9.23)
Tail Risk
0.82
100%
★ NEW v3.4 - VIX/MOVE proxy (t=-3.23)
Geopolitical
0.56
99.4%
★ NEW v3.4 - GPR proxy (t=2.54)
Weekend Effect
0.53 (2x)
100%
Fri→Mon mean reversion (2x during active)
Momentum
0.18
99.1%
1-day lag mean reversion (t=-2.32)
Flow
0.19
88.5%
ETF flows (t=-1.21)
Macro
0.08
83.0%
DXY correlation (t=0.96)
Google Trends
0.03
82.2%
Attention proxy
Supply-Side
0.01
57.5%
Negligible (t=0.18)
Derivatives
0.01
55.8%
Negligible (t=-0.15)
Regulatory
0.05
N/A
Qualitative
v3.4.0: Wild Bootstrap OLS (40k iterations) + 5-fold CV. NEW: Geopolitical (99.4% sig) + Tail Risk (100% sig) for crisis defense. PhD-validated HAC standard errors.
Known Limitations
- Weekend Effect only active Fri 4pm CT → Mon 9:30am ET (CME hours)
- CME planning 24/7 trading in 2026 may reduce weekend effect
- Macro & Flow have reduced significance in parsimonious model
- ETF flow data only available from Jan 2024
- Model underperforms in range-bound/choppy markets
Signal Formula
SIGNAL = Weighted Avg of Category Scores × Regime Factor × 0.95
Probability via logistic transformation: P = 1/(1+exp(-0.8×signal)), bounded [30%, 85%]
Wild Bootstrap OLS: May 2020 - Jan 2026 (2,071 obs) | 10,000 iterations | R²=0.125, F=21.36 | Weekend Effect: 100% sig