Daily returns or range for RV?

[MrAgi1]

Premium sellers would claim that “implied volatility(I.V) overstates realized volatility(RV) over the long term”, so the edge of selling options is the risk premium.

As we know, volatility can be calculated in different ways. RV is calculated using daily returns(i.e |open-close|). What if instead RV is calculated using the trading range(i.e high-low) just like the A.T.R:

1. Using the trading range(i.e high-low) to calculate RV, would IV still overstate RV in the long term?

2. Why is RV not calculated using the high minus low method?

3. Would managing option buying strategies by setting intraday targets(rather than waiting for the close) affect RV’s long term profitability against IV?

 

[Sekiyo]

RV is calculated using daily returns

Are you really sure about that ?
Sounds theoretically and practically weird.

Change is "close-open" and not the other way around

For example, "20-day historical volatility" measures realized volatility over last 20 days (it is typically calculated as standard deviation of last 20 daily price changes).

Input sounds to be Change ...

Using daily range would make more sense to me.

Anyway.

 

[MrAgi1]

Are you really sure about that ?
Sounds theoretically and practically weird.

Change is "close-open" and not the other way around

For example, "20-day historical volatility" measures realized volatility over last 20 days (it is typically calculated as standard deviation of last 20 daily price changes).

Input sounds to be Change ...

Using daily range would make more sense to me.

Anyway.

Your are correct. It should be close-open not the other way.

However it still looks weird to me why trading range is not input, instead change is used to calculate the RV.

standard deviation of the trading range(i.e high-low) makes more sense.

 

[Baozi]

There are many ways to estimate RV.
google garman klass or yang zhang..

 

[Baozi]

also, I don't think using the ATR you could compare it directly to IV, that would be like comparing apples to oranges..

 

[newwurldmn]

We use close to close mostly because it’s how most people trade.

There are realized vol models that use high to low or open as part of their vol calculation. Garman klas is a famous one.

you can use any calculation you want provided you can trade at that vol (ie replicate the option hedging at those prices).

Premium sellers would claim that “implied volatility(I.V) overstates realized volatility(RV) over the long term”, so the edge of selling options is the risk premium.

As we know, volatility can be calculated in different ways. RV is calculated using daily returns(i.e |open-close|). What if instead RV is calculated using the trading range(i.e high-low) just like the A.T.R:

1. Using the trading range(i.e high-low) to calculate RV, would IV still overstate RV in the long term?

2. Why is RV not calculated using the high minus low method?

3. Would managing option buying strategies by setting intraday targets(rather than waiting for the close) affect RV’s long term profitability against IV?

 

[Matt_ORATS]

We provide both close to close and a modified Parkinson.

When I was a floor trader, we devised a method using stock tick data to simulate hedging gamma and turn that profit into a volatility. We concluded that this method was the closest to mirroring what went on with actual trading. For example, if we bought IV at 50% a realized vol of 54% should have us making money with the position. This was a real world case with Nextel where close to close said 50% but we were making money with a 50% IV. Later, after we developed the above method, the tick vol was 54%.

We were able to modify the Parkinson HV (OrHV) to match our tick volatility and that is what we use today.

We provide forecasts using our OrHV and test them against IV as a forecast.

For ~1700 liquid stocks, our r-squared for the independent variable of the forecast of thnext 20 observations with a dependent variable of the actual 20 day volatility over a one year period is 0.32 and the rsq for IV is 0.27. This has been consistent over time. We usually beat the IV rsq by 0.05 or so and the overall rsq is around 0.30.

The top 10 rsq OrHv with rsq>0.55 are:
ticker
AR
MSTR
BTU
DM
UNG
ARKK
GNRC
SHOP
AXL
RCL

The bottom 10 with rsq< 0.04 are:
DNA
ZIM
LI
ICLN
NU
DWAC
SOFI
GOLD
SNOW
BITO

Interesting BITO is the worst and we show a large edge of IV - HV.

 

[newwurldmn]

We provide both close to close and a modified Parkinson.

When I was a floor trader, we devised a method using stock tick data to simulate hedging gamma and turn that profit into a volatility. We concluded that this method was the closest to mirroring what went on with actual trading. For example, if we bought IV at 50% a realized vol of 54% should have us making money with the position. This was a real world case with Nextel where close to close said 50% but we were making money with a 50% IV. Later, after we developed the above method, the tick vol was 54%.

We were able to modify the Parkinson HV (OrHV) to match our tick volatility and that is what we use today.

We provide forecasts using our OrHV and test them against IV as a forecast.

For ~1700 liquid stocks, our r-squared for the independent variable of the forecast of thnext 20 observations with a dependent variable of the actual 20 day volatility over a one year period is 0.32 and the rsq for IV is 0.27. This has been consistent over time. We usually beat the IV rsq by 0.05 or so and the overall rsq is around 0.30.

The top 10 rsq OrHv with rsq>0.55 are:
ticker
AR
MSTR
BTU
DM
UNG
ARKK
GNRC
SHOP
AXL
RCL

The bottom 10 with rsq< 0.04 are:
DNA
ZIM
LI
ICLN
NU
DWAC
SOFI
GOLD
SNOW
BITO

Interesting BITO is the worst and we show a large edge of IV - HV.

I remember DNA being Genentech. That was a fun risk arb opportunity.
Everyone on CNBC Fast Money became an expert on Cholorectal cancer over night.