So im at this stage of my trading path where I’m familiar with the concepts of volatility. Yet I don’t know how to value it.
for example I see advanced traders always saying you need a “model” that can value etermine if it’s cheap/expensive..
There are different kinds of model.
There are pricing models, like Black-Scholes. These can translate volatility quoted in different units, so for example given an option price in $ and some other inputs you can work out the "implied volatility". You can also do this calculation in the other direction, so if you have an idea of what implied vol should be for a particular option you can calculate a fair value for the price. Most option pricing platforms do this for you.
None of this is telling you whether a particular option is cheap or expensive, since you still need a 'fair value' for the implied vol. To do this you need to compare the implied vol of the option to other options, or make a forecast of what you think will happen to volatility (note we always do these comparisions in implied vol space, not price space).
A simple way to compare an option to other options is to construct a 'volatility surface'* (confusingly this approach is sometimes called 'model free' since it makes no assumptions about the process driving prices). When you do this you may find options that are above or below the surface and therefore cheap or expensive.
* 3-d for most instruments, and 4-d for interest rate products where the maturity of the interest rate is a factor
For example, for a given duration of option maturity and given instrument, the surface is just a 2-d graph with strike on the x-axis and implied vol on the y-axis. If you plot all the implied volatility points on this graph, you can try and draw a line through them using something like a quartic spline (basically a curvy line- don't panic,
Excel can do this). Don't use a straight line, as this won't account for the 'smile' or 'skew' of option prices, and you'll always be selling out of the money options.
* Note: Make sure you use a spline with fewer 'degrees of freedom' than you have points. Otherwise the line will go through all the points and be useless. For example if you have 5 strikes, then don't use anything more complex than a cubic polynomial.
If you find some points above this line then these are expensive options you should sell, and vice versa. Often these points will be out of the money vols where the wide bid-ask spread means you can't actually exploit this apparent mispricing (to check this, translate both the bid and the ask prices into implied volatility terms).
Alternatively, you can compare the volatility surface to the historic shape of the surface. Again for the simple 2-d example you would compare the current implied vols to an estimate of the curvy 'fair value' line based on historic data (typically you would shift the historic line up or down so that it went exactly through the at the money strike, or it would look like all strikes were undervalued or overvalued). Intuitively then, if the 'smile' or 'skew' of options is way out of line with historic levels you would end up buying or selling away from the at the money.
You can also try and forecast what volatility will do in the future ("realised volatility"). If the implied vol is higher than the expected realised vol then we'd sell options and vice versa (ideally we'd be selling straddles or strangles and delta hedging, but you can still use this approach to give you an idea of whether options are currently overvalued or undervalued).
A simple way of doing this is to assume that future realised vol will be roughly what it averaged in the past few years ("mean reversion"), and if implied vol is higher than that you sell vol, and vice versa. Because mean reversion doesn't happen that quickly this works best on longer dated options (at least 3 to 6 months).
There are various more complex ways of doing this, such as GARCH (mentioned by
@TheBigShort). These are all models where you estimate the model parameters based on past data, and then extrapolate from the most recent levels of vol to forecast what it will be in the near future (again as
@TheBigShort says you can make these models more accurate by using higher frequency data to estimate recent levels of volatility). More sophisticated models will take into account things like economic announcements and fed meetings that will briefly bump up volatility.
Hope this helps.
GAT
