Thread by @WizOfOps

This tweet was my most viral tweet, and I talked about deviation from that relationship being scenarios where MMs become "uncomfortable". Why do they become uncomfortable?

In short, this means that their assumptions surrounding skew have been wrong. 1/13

https://twitter.com/WizOfOps/status/1446117113735749638

A commonly used term for the "vixing" relationship on Twitter and among quants is "spot-vol covariance". This relationship isn't novel, it isn't something no one knew before I tweeted it. As such, this is the assumption that creates skew. See that formula on the regression? 2/13

Every 1% move in $SPX is associated with a move in the $VIX. That association is reflected in the skew curve. Start with 0% move, then move up 1%, and the covariance of the hour determines the IV at that strike. Work all the way to a 0 bid, voila, you have assumed skew. 3/13

Here's a few things to note. $VIX isn't used in the "over/under" vixing calculation for skew. "Fixed-price vol" is used, which basically says "what is vol when we reach the skewed strike, more or less than what we predicted?" VIX however is a needed convenient proxy. 4/13

When the IV in that skew is higher than what was predicted, we are "overvixed". I'm sure in some back room, a bunch of quants get together and discuss what the real fair value is, and make the market towards their new valuation. Historically, it isn't too different. 5/13

And that's the machinations of what makes "overvixing" work. Now someone asked why VIX has been underperforming, or sometimes performing poorly compared to the vixing. In truth, it isn't. This whole year we have not overvixed once more than 2.5 points. 6/13

That doesn't mean this doesn't work, it just means that this year MMs have done an excellent job handling premiums as they relate to skew.

Many have been talking about how skew is flat. Historically it is, the last 6mo regression formula is lower than normal... 7/13

But for now it is working in accordance with the predicted skew. TBH, my thesis is that the excessive 0DTE traffic has something to do with it. I can't prove that yet, but 0DTE options are a quick, cheap way to hedge risk. Gamma, vanna, and charm are more pronounced... 8/13

and allows for better risk management. That doesn't mean overvixing is over, it just means that speculation needs to get very one-sided to create a spike in IV, where no one wants to take the other side of 0DTE hedging activity. 9/13

How one-sided are we? Looking at #Volland, here are $SPY and $SPX outstanding deltas. These are hedged, so there's no action, but just see how varied they are. These are the 2 largest vehicles in average daily notional option volume. Negative in SPY, positive in SPX. 10/13

Then 1-sided action in individual equities for the most part... creating the volatility we are seeing now. But as long as there are 2-sided markets in options, we aren't going to see a major $VIX explosion. The more 2-sided we are, the better the "vixing" assumption. 11/13

When you see modest movements in the vixing, like $VIX up, $SPX up or vice versa (sans Mon&Fri), you need to see how that relates to the skew, and if there is a repricing of $VIX upcoming, or if it is event vol, or what is happening. If the answer is nothing foreseeable... 12/13

you will likely see a repricing of spot/IV towards the previous vixing assumption in the skew. If there is something to it, it may not be a "reversal" signal.

So when I make directional calls based on IV that makes everyone question me, it is usually based on this dynamic. 13/13