CASE 04 · VOLATILITY RESEARCH

Volatility Timing for Short-Volatility Carry

Heston–UKF filtering · Implied-realized spreads · Dynamic short-strangle allocation

UNDERLYINGS2 ASSETS

SPY and AAPL option chains.

STRATEGIES3 VARIANTS

Static, Heston–UKF and RV_21d timing.

CALIBRATION42 DAYS

Rolling window with a 10-day refit schedule.

EXPOSURE0×–2×

Clipped short-volatility allocation.

RESEARCH QUESTION

Can volatility estimates improve carry allocation?

The option trade remains a short one-week strangle. Only its exposure changes with the spread between market-implied volatility and estimated realized volatility.

Estimate

Filtered latent Heston variance from daily returns with an Unscented Kalman Filter.

Compare

Measured implied volatility against UKF and 21-day realized-volatility estimates.

Scale

Smoothed and standardized each spread into a clipped 0×–2× allocation.

Backtest

Applied the allocation to the same short one-week strangle strategy.

Benchmark

Compared total return, Sharpe, Calmar and drawdown on SPY and AAPL.

EMPIRICAL EVIDENCE

Timing helps; complexity does not automatically win.

Both timing signals improved total return over static carry, while the simpler RV_21d rule remained the strongest specification in these samples.

SPY volatility proxies, implied-realized spreads and dynamic allocations — FIGURE 01 · SIGNAL PIPELINESPY volatility estimates and dynamic allocation · 2020–2022
The UKF and RV_21d spreads produce distinct allocation paths under the same 0×–2× exposure constraints.

SPY short-volatility carry NAV comparison — FIGURE 02 · SPY3.81% static · 4.51% UKF · 5.20% RV_21d
UKF timing improved total return, but the simpler RV_21d signal delivered the highest Sharpe at 0.855 and the lowest drawdown at 1.77%.

AAPL short-volatility carry NAV comparison — FIGURE 03 · AAPL ROBUSTNESS18.66% static · 26.01% UKF · 28.99% RV_21d
Both timing rules improved the static strategy over the 2016–2023 sample; RV_21d again ranked first.

CORE FINDING

Allocation matters more than model complexity.

The Heston–UKF signal was directionally useful on both assets, yet a transparent 21-day realized-volatility benchmark produced stronger risk-adjusted results.

LIMIT

Backtest evidence, not live performance.

Production use would require explicit transaction costs, margin, option-surface liquidity, leg execution and stability tests beyond two underlyings.