Research & Insights
Original research from the svadri.ai team — spanning factor dynamics, alternative data, machine learning in finance, and risk modeling. All data and results are simulated unless otherwise noted.
We document a systematic decay in cross-sectional momentum alpha when intraday volatility exceeds the 90th percentile. Our regime-adaptive model adjusts factor weights dynamically, recovering 34% of lost alpha during high-volatility episodes.
Using transformer-based sentiment scoring on 12 years of earnings transcripts, we identify a statistically significant 3-day drift following tone shifts in management guidance. The signal is orthogonal to traditional price momentum.
Traditional VaR models underestimate tail risk by 40–60% during regime transitions. We propose an EVT-augmented framework that captures fat-tail dynamics more accurately, with out-of-sample validation across three market crises.
Most quantitative signals exploit correlation without establishing causation. We introduce a causal inference framework — grounded in do-calculus — that filters spurious signals and improves out-of-sample Sharpe by 0.4 on average.
We apply Hidden Markov Models to classify macro regimes — trending, mean-reverting, and crisis — across 30 years of multi-asset data. Regime-aware position sizing reduces maximum drawdown by 28% with minimal alpha sacrifice.
Alternative data signals degrade faster than traditional factors as adoption spreads. We model the half-life of alpha across 14 alternative data categories and propose a rotation framework that systematically harvests early-adopter edge.
New papers and research notes published monthly. Join our distribution list for early access.
Request Research Access