- Joined
- 7/1/24
- Messages
- 1
- Points
- 1
Hi,
I want to create a volatility arbitrage screener for finding the best trade at time t. Let's define S as the underlying contract, V as a vanilla option on that underlying, T as the option maturity, and t as the current moment. I want to exploit the variance risk premium, which is the difference between the Black-Scholes (BS) implied volatility of the option at time t and the realized volatility of the underlying contract between t and T (option maturity). We assume delta hedging with the underlying so the portfolio delta is close to neutral.
I want to use the Parkinson estimator for calculating historical volatility. If we use N=T (we can suppose T=30 days), we calculate the volatility for the last 30 days. However, we can't use this value for the forecast.
We need to forecast the volatility over the next 30 days, and the latest volatility can give us the forecast for 30-day volatility at time t=1, if we assume for simplification that historical volatility will be the future volatility – for the next period only.
I was thinking about resampling the data to match the DTE (days to expiration) period. Concretely, if the option expires in 5 days, we can use weekly sampled price data (1 OHLC per week). We can extend this reasoning to N days for every maturity.
Do you think this approach is correct? I have seen something similar in Artur Sepp's "Volatility Modelling and Trading," but Sheldon Natenberg's approach in his book is contrary to the above logic.
Thanks
I want to create a volatility arbitrage screener for finding the best trade at time t. Let's define S as the underlying contract, V as a vanilla option on that underlying, T as the option maturity, and t as the current moment. I want to exploit the variance risk premium, which is the difference between the Black-Scholes (BS) implied volatility of the option at time t and the realized volatility of the underlying contract between t and T (option maturity). We assume delta hedging with the underlying so the portfolio delta is close to neutral.
I want to use the Parkinson estimator for calculating historical volatility. If we use N=T (we can suppose T=30 days), we calculate the volatility for the last 30 days. However, we can't use this value for the forecast.
We need to forecast the volatility over the next 30 days, and the latest volatility can give us the forecast for 30-day volatility at time t=1, if we assume for simplification that historical volatility will be the future volatility – for the next period only.
I was thinking about resampling the data to match the DTE (days to expiration) period. Concretely, if the option expires in 5 days, we can use weekly sampled price data (1 OHLC per week). We can extend this reasoning to N days for every maturity.
Do you think this approach is correct? I have seen something similar in Artur Sepp's "Volatility Modelling and Trading," but Sheldon Natenberg's approach in his book is contrary to the above logic.
Thanks
