VolVantage — NIFTY Volatility & Options Liquidity Engine

Select Asset / Index

Lookback (Days): 45

Top Model: LightGBM

Best QLIKE: 0.00841

Out-of-Sample R²: 0.6842

🎯 Out-of-Sample Forecast vs Realized Volatility

Comparing ground truth Realized Volatility (%) against Econometric HAR variants and Machine Learning models

Model Performance Leaderboard

Ranked by Patton (2011) robust Quasi-Likelihood (QLIKE) loss

Rank	Model	QLIKE Loss	R² Score	RMSE

Diebold-Mariano Significance Matrix

Pairwise loss differential tests. Cells marked with ★ indicate p < 0.05 statistical significance

Realized Volatility & Jump Variation Decomposition

Separates total integrated variance into Continuous Diffusion (Bipower Variation $BV_t$) and Jump shocks ($J_t = \max(0, RV_t - BV_t)$)

Index / Underwriting Name

Target DTE (Days to Expiry)

Min Liquidity Score: 30

Rank Score = 40% Liquidity + 35% Safety + 25% Yield

⚡ Ranked Short-Strangle Setups

Click any row to dynamically load its live P&L payoff curve and diagnostic profile

25 Candidates

Expiry	DTE	Put Strike (PE)	Call Strike (CE)	Credit (₹)	Yield %	Net Delta	Mean IV	Liquidity	Safety	Rank Score

📊 Expiration P&L Payoff Diagram

Selected Strangle: Put 24,000 / Call 25,000

Lower BE: ₹23,840 Upper BE: ₹25,160

Position Diagnostics

Spot Price: ₹24,680.00

Total Credit Collected: ₹160.00

Net Directional Delta: +0.0120

Strangle Width (Cushion): 4.05%

Composite Safety Score: 82.5 / 100

Open Interest Depth: 284,500 lots

⚡ Execution Tip: High-probability delta-neutral strangle setup with wide breakeven cushion.

Intraday OI Divergence vs Price Trajectory

Flags asymmetric Open Interest writing against spot direction (resistance/support buildup)

Option Chain OI Depth Ladder

Call Open Interest (Red) vs Put Open Interest (Green) across strikes

🏛️ Mathematical Formulations & System Architecture

VolVantage is engineered to run high-throughput quantitative volatility forecasting and options liquidity analytics on embedded DuckDB without requiring an external database cluster.

1. Realized Volatility & Jump Decomposition

From 5-minute intraday log returns $r_{t,i} = \ln(P_{t,i}/P_{t,i-1})$:

$$RV_t = \sum_{i=1}^M r_{t,i}^2, \qquad BV_t = \frac{\pi}{2}\sum_{i=2}^M |r_{t,i}||r_{t,i-1}|, \qquad J_t = \max(0, RV_t - BV_t)$$

2. HAR Cascade & Spillover Extensions

The Heterogeneous Autoregressive model incorporates multi-scale cascade memory and cross-asset correlation/sector spillover factors:

$$RV_{t+1, i} = \beta_0 + \beta_d RV_{t,i} + \beta_w RV_{t,i}^{(w)} + \beta_m RV_{t,i}^{(m)} + \gamma_c \overline{RV}_{t, \text{cluster}(i)} + \varepsilon_{t+1,i}$$

3. Patton (2011) Robust QLIKE Loss & Diebold-Mariano Tests

Standard MSE is asymmetric and non-robust to heteroskedasticity in volatility modeling. We evaluate models using the Quasi-Likelihood robust loss:

$$\mathcal{L}_{QLIKE}(y_t, \hat{y}_t) = \frac{y_t}{\hat{y}_t} - \ln\left(\frac{y_t}{\hat{y}_t}\right) - 1, \qquad DM = \frac{\bar{d}}{\sqrt{\hat{V}(\bar{d})}} \sim \mathcal{N}(0, 1)$$

4. Short-Strangle Multi-Factor Ranking Function

Candidates are prioritized by balancing liquidity depth against portfolio risk and premium yield:

$$\text{Rank Score} = 0.40 \cdot \mathcal{S}_{\text{Liquidity}} + 0.35 \cdot \mathcal{S}_{\text{Risk Safety}} + 0.25 \cdot \mathcal{S}_{\text{Yield}}$$