If “high” was expensive and “low” was cheap then trading would be easy. I’ve discussed this tension in:
I was recently asked the following question:
Hey Kris, got a beginner question for you if you have some time. Why do people recommend selling ATM spreads instead of slight OTM? If there’s a smile, it seems to make sense to sell the higher IV wing.
Re-stated more generally, the question is:
Why would I ever buy a higher IV or “skewed” option to sell a lower IV option?
You can get into a long discussion about greeks, liquidity, jump probabilities, distributions and their moments, and spot-vol correlations. They will all lead you back to the idea that “high IV” doesn’t mean expensive IV. It’s not an encouraging answer if you are looking for simplicity.
But let me offer a constructive perspective to help you along.
It’s not hard to understand why skew exists in option markets:
- supply/demand of risk (ie hedging and overwriting flows)
- correlations increasing when risk premiums expand (here’s my thread on dispersion)
- fundamentally, a stock is more levered when its equity value falls
In addition to those, I’m sure there are technical (ie lots of math) reasons involving jump models and higher statistical moments. I’m not smart enough for that. Many option traders probably aren’t. But one of the ways to survive/thrive is to take a more intuitive approach.
The logic flows as follows:
- Markets are pretty smart. It’s naive to think “high” equals expensive.
- Implied vols are a useful ruler for comparing vols but I can’t read too much into them as valuation tools since the underlying distributions are unknowable.
- Market prices contain extra intelligence or assumptions about a stock’s distribution but Black-Scholes assumes a singular distribution leading to differential implied vols. (Those differences are a fudge because we are standardizing the underlying distribution, even though we know the market is capable of handicapping a multitude of conditional distributions.)
- Focus on relative pricing to make your process less model-dependent. This lets the model errors “cancel out”.
Here’s an example of this relative thinking that I explained to the learner:
Suppose I found 20 reasonably correlated names that all have skews more expensive than at-the-money IV. If you sorted the skewed options as a percent of ATM vol there would be a top half and bottom half of expensiveness. But if you looked at just the cheapest one naively in isolation you would want to sell the skewed option. But zoomed out in a cross-sectional view you would have wanted to buy it.
If you are only trading one name you are in the domain of my post Structuring Directional Option Trades. In this case, your fundamental analysis is upstream of your option trade expression. So be careful about mixing up vol trading which requires a zoomed out lens and directional options trading which requires a deep understanding of a single name’s distribution (see Real Talk On Options Trading).
A possible compromise between the approaches is to look at a time series of the skew relative to ATM to see if it’s low end or high end of normal. This will still deceive you in cases when all the skews in the market converge. For example when all skews in the market are “high”, if you look at your name in isolation you still won’t know if it’s relatively “high”. A proper cross-sectional method will benchmark to a liquid name or basket that can be considered “fair”.
So that’s 3 lenses. Cross sectional, fundamental, and time series. It would be nice if your trade idea looked good on all 3 filters, but option traders usually have limited visibility into fundamentals so it’s too high a bar for pulling a trigger.
How can option traders make up for that incomplete picture? The same way poker players use betting patterns to narrow a hand.
Here’s a clue:
