About Aegis Finance
Methodology, chart reading guide, known limitations, and credits
Methodology
Jump-diffusion model (Merton 1976) with GJR-GARCH(1,1) stochastic volatility and 3-state Hidden Markov Model regime blending. 10,000 paths across 7 probability-weighted scenarios.
- -Variance drag correctly accounted for (geometric drift, no double-counting Ito correction)
- -Merton jump compensator ensures E[S_T] = S0 * exp(mu * T) even with jumps
- -Return cap at 300% over 5 years prevents unrealistic outliers
- -Mean reversion activated when prices deviate 20%+ from fair value
LightGBM + Logistic Regression ensemble (70/30 blend) with isotonic calibration. 25-30 features selected via LASSO from 208 candidates.
- -3-month, 6-month, and 12-month horizons
- -Walk-forward backtesting with zero data leakage
- -Target: Brier Score ≤ 0.05 (random = 0.25, climatology ~0.12)
- -Crash defined as 20%+ drawdown from recent peak
9-factor composite z-score combining VIX, yield curve, credit spreads, momentum exhaustion, short-term volatility, gold/stock ratio, market breadth, and small-cap divergence.
- -Range: -4 (very low risk) to +4 (extreme risk)
- -Above 2.0 = elevated stress, triggers scenario weight adjustments
- -Factors weighted by historical predictive power
Multi-check regime classification (Bull/Bear/Volatile/Neutral) validated against 200-day SMA, market breadth, and institutional consensus.
- -Prevents false bear signals during temporary corrections
- -Regime feeds into Monte Carlo scenario weights and drift adjustments
How to Read the Charts
Fan Charts
Width shows uncertainty. The blue line is the median (50th percentile). Shaded bands show 5th-25th-75th-95th percentile ranges. Wider bands = more uncertainty.
SHAP Waterfall
Red bars push crash probability UP (more risk). Green bars push it DOWN (less risk). Length shows magnitude of impact. Features sorted by absolute importance.
Sector Heatmap
Color intensity shows expected return magnitude. Greener = higher expected return. Red = negative expected return. Based on Monte Carlo with sector-specific parameters.
Crash Gauges
Semicircle gauge showing crash probability. Green (0-15%) = low risk, Yellow (15-30%) = moderate, Amber (30-50%) = elevated, Red (50%+) = high. Probabilities should increase with time horizon (3m < 6m < 12m).
Correlation Matrix
Red (>0.7) = highly correlated (move together, less diversification). Green (0-0.4) = moderate. Blue (<0) = negatively correlated (good for diversification).
Known Limitations
- -All predictions are probabilistic estimates with significant uncertainty
- -Past performance does not guarantee future results
- -Model assumes stationarity that may not hold during structural breaks
- -Crash prediction has limited positive predictive power (rare events are inherently hard to predict)
- -Yahoo Finance data may have gaps, especially for smaller or international tickers
- -FRED data updates with lag (some series are monthly or quarterly)
- -Jump-diffusion model cannot capture all market dynamics (correlations, sector contagion)
- -No transaction costs, taxes, or slippage modeled
Data Sources
Yahoo Finance
Prices, VIX, sector ETFs, treasuries, analyst targets, fundamentals. Updated hourly (cached).
FRED (Federal Reserve)
22+ macro series: yield curve, unemployment, CPI, NFCI, initial claims, LEI, SLOOS. Cached 24hr.
GDELT
Global event tone, volume, conflict scores for news intelligence. On request.
Credits
Built by Murathan as a research project combining quantitative finance with modern web development.
Open source under the MIT License.
Disclaimer
Aegis Finance is an educational tool. It is not financial advice. All predictions are probabilistic estimates with significant uncertainty. Past performance does not guarantee future results. Always consult a qualified financial advisor before making investment decisions.