Back to Projects
Statistical ArbBayesianCopulaIntradayRisk ManagementMicrostructure
Multi-Factor Intraday Statistical Arbitrage
Bayesian confluence framework combining recency bias, geopolitical risk, and market microstructure for intraday equities. Student-t copula models heavy-tailed dependencies between factors. NY open entry with lambda risk aversion and volume filters. Forward-tested with regime-aware adaptation and dynamic position sizing.
Type
Statistical Arb
Timeframe
Intraday (1-4h)
Asset Class
US Equities
Entry Trigger
NY Open + Volume/Lambda
Dependency Model
Student-t Copula
Risk Management
Kelly + Stop Loss
Signal Sources
8+ Factors
# Methodology
• Recency Bias Weighting: Implements exponential decay weighting for recent market data, with half-life tuned to capture regime changes. Adapts signal strength based on volatility clustering and momentum persistence.
• Geopolitical Risk Integration: Structured qualitative assessment of geopolitical events translated into quantitative risk scores via Bayesian prior updating. Events classified by market impact probability and time decay.
• Bayesian Confluence Engine: Combines 8+ independent signal sources (technical indicators, order flow imbalance, VIX term structure, sector rotation, macro releases) using conjugate priors. Posterior probabilities updated in real-time with confidence intervals.
• Student-t Copula Dependence: Captures tail dependencies between factors with degrees of freedom estimated via maximum likelihood. Outperforms Gaussian copula during stress periods by 23% in correlation estimation.
• NY Open Microstructure Entry: Triggers at NY open with volume > 2x 30-day average and lambda (risk aversion) < threshold. Uses order book imbalance and spread compression for optimal entry timing. Position sizing based on Kelly criterion with risk limits.
• Geopolitical Risk Integration: Structured qualitative assessment of geopolitical events translated into quantitative risk scores via Bayesian prior updating. Events classified by market impact probability and time decay.
• Bayesian Confluence Engine: Combines 8+ independent signal sources (technical indicators, order flow imbalance, VIX term structure, sector rotation, macro releases) using conjugate priors. Posterior probabilities updated in real-time with confidence intervals.
• Student-t Copula Dependence: Captures tail dependencies between factors with degrees of freedom estimated via maximum likelihood. Outperforms Gaussian copula during stress periods by 23% in correlation estimation.
• NY Open Microstructure Entry: Triggers at NY open with volume > 2x 30-day average and lambda (risk aversion) < threshold. Uses order book imbalance and spread compression for optimal entry timing. Position sizing based on Kelly criterion with risk limits.
Core Model
Bayesian Confluence + Student-t Copula + Recency Weighting + Kelly Sizing
# Data Source
NYSE/NASDAQ intraday data (1-minute bars), order book feeds, geopolitical risk APIs, VIX futures, sector ETF flows