i see. to hedge the vol risk you need to spend some of your capital to put on another similar (meaning long gamma) strategy that profits from a rising IV, and rehedge whenever the portfolio's Vega gets above a certain absolute threshold (+ve or -ve). A straddle is one simple example. I'd recommend this over using VIX derivatives because the correlation, as others have mentioned, may be far from perfect, and you might not get the hedge when you most need it. On your 2nd answer, yes there's an easy way to model different underlyings together in TOS. In the analyze page just click on "Portfolio, beta weighted" instead of "Single". Hope this helps.
@rallymode, hlpsg Thanks for your replies! A straddle is one simple example. I'd recommend this over using VIX derivatives because the correlation, as others have mentioned, may be far from perfect, and you might not get the hedge when you most need it. I backtested hedging the vega with a SPX straddle and gamma scalping it so that I can eliminate theta and have only vega left. But this hedge (gamma scalping a straddle) is also not perfect and anyway it can result in losses when IV rises :-( Other question: How are the pathways of VIX and S&P500 linked? Eg. if SPX falls 10 points where will VIX be? Or SPX falls 1% how much percent will VIX approximately/normally rise? Thanks and regards
There is very high correlation, but correlation doesn't tell you the % change given 1% change in SPX. And given the mean-reverting nature of volatility I don't think you can make any conclusions with respect to the SPX 1% - VIX x% relationship. Your main problem is that VIX is a poor proxy for the implied volatility of the option you have, and not the relationship between SPX and VIX.
what were your total portfolio Vegas (reverse calendar + straddle) when you say you lost money when IV rises? Vega is not static, if my observations are correct, it has a "delta" too and the first derivative of Vega (i.e. the delta of Vega) is probably best described as a "U" shaped curve if I'm not mistaken. So you always need to rehedge your Vega as IV rises or falls above a certain extent.
Going beyond Vega you have Vanna (sensitivity of Vega to changes in price of the underlying) and Vomma (sensitivity of Vega to changes in volatility).
Once you filter out synthetic time there is a very well defined and quantifiable inverse corr within 1 sigma[monthly] though it's better if you measure in points or handles vs %. Regardless, i would not suggest hedging SP vol with forward vol when the term structure is in contango. Better to replicate elsewhere.
@Rallymode Once you filter out synthetic time there is a very well defined and quantifiable inverse corr within 1 sigma[monthly] though it's better if you measure in points or handles vs %. Is that calculation open to the public and can you describe its calculation? So when market moves morre than 1 sigma the caluclation is no more reliable? Thanks
@rallymode I found in the www an approximation for an expected VIX move when SPX falls 10%: a 10% SPX down move comes with a 50% VIX rise. Is this a good approximation or not cause it always depends on time and one sigma? Regards and thanks