Thread by @TerribleQuant

I'm blatantly copying and pasting this from Mephisto but the fact that some people haven't seen this thread, and will otherwise never be able to since the account is gone is sad:

OK, picking apart $SKEW. Why is it 'hot garbage' -h/t @ jsoloff. This is for anyone who read @ SoberLook (who of course are awesome but the problem is elsewhere)

"Have you met my bro $SKEW? He's useless but his dad owns a boat."

1/ too many

Now the thing is, as always, that the exchange are making a fortune from fees on $VIX. But it's their only baby and they'd love another one just in case something happens to this one.
2/

So, a while back, I guess in a committee meeting that should have been an email, someone suggested 'ooh ooh lets do skew next'. Of course, quants had been there already, there's a replicating portfolio for the third moment just like the second. BINGO.
3/

So, just like in the $VIX whitepaper, they take the integral over strikes, discretize it, add some cutoff rules, and da-da there's your $SKEW index. The whitepaper basically writes itself. (In fact it looks a lot like an intern did)
4/

But there's a problem. $VIX is vol which is like noise which they figured out how to sell as the 'fear gauge'. Brilliant marketing. Absolute nonsense, but the muppets applaud, CNBC gets a chart for Markets in Turmoil, 50c lifts 300k $VIX options each month, everyone's happy
5/

Ok for the 3rd moment what can we do. Committee ponders... it's the "demand for tail puts", "higher power of fear"... they're struggling and you can see the problem. Once youve successfully sold $VIX as the fear gauge you cant go to a higher power and go 'ABSOLUTE PANIC GAUGE'
6/

So they dick around with it but kind of lives and kind of dies. There's no futures on it. However, it lives on in published materials and in Bloomberg as "SKEW INDEX" which is always going to sucker some journalists, students, and tourists. (Sorry, not sorry.)
7/

So basically CBOE give up. They don't know what to do with it. But its easy to calculate, it gets some airtime, so they keep it. It's another in the long line of $VIX failsons, the Chad Laxbro of vol indices. Have you met my bro $SKEW? He's useless but his dad owns a boat.
8/

And there's another, bigger problem. It's junk. The timeseries is just a mess. take a look at that chart again. It's the same number in May 2020 and in Octobor 2017, but goes to all other numbers from 120 to 145 randomly in between.
9/

You've sold it as "fear squared" but it can collapse when the market does - Q4 2018 for example. It does nothing at all in Feb18, one of the single biggest 1-day shocks ever, and then stays low all 2019. Why? Fuck knows. You couldn't screw up an indicator harder if you tried.
10/

Ok so now lets dig a bit. The $SKEW whitepaper has the answers so I don't need to do any algebra here. Blah blah 3rd moment, cubes, expectations, lets say it's all done and ticked off. Relief all round
11/

Crack open that whitepaper go to chart 2, remembering those 120-145 multi-year ranges. Now check out that red and orange line. There's a 12+ point difference in the $SKEW valuation but the put wing is near identical. Ooops. You realise now what's going on.
12/

That the 3rd moment is anti-symmetric - its like "puts-calls" (with weights to puts), whereas $VIX is "puts+calls" (with weights to puts). But the calls still impact the results, and the - sign means the result is small so you have to scale up, scaling the noise too.
13/

The call wing clearly can, and does move $SKEW 12 points on it's own with the puts completely unchanged. That's half of your multi-year 120-145 range. You're clearly fucked.
14/

This isn't a tail risk index or a fear gauge. This is a quant number. This really is the third moment of the distribution. You've put a huge weight on teeny tiny premiums on both sides - and in the real world it isn't so pretty, it's just too noisy.
15/

More details appear in Table 12 after 15 pages of formulas. In Microsoft Word. Good God, the intern really did write this - they must have given up early. Anyway, you can see the negative weighting toward 10 cent calls, versus the positive weighting to 5c puts. Awesome.
16/

For those of you under 30 - yes the SPX was at 1105 once, and we all sold it there because it was clearly TOO HIGH. Man, I'm old.
17/

You can also see that the cutoff rule will have more impact here. CBOE stop the sum over strikes of $VIX and $SKEW when you hit 2 zero bids in a row. This is kind of OK for $VIX because the meat of the $VIX components is at tighter strikes, where zero bids don't happen.
18/

But $SKEW weights the really wide tails much more as it's a higher moment. So, say Socgen's calc engine b65df3h@efd says their exo desk is painfully long 115% strike 1 month calls and they hit every nickel bid they can find.
19/

Server 14bXfq!73 at Citadel pull the bids on 2 of the consecutive strike calls on SPX that nobody else cares about, and $SKEW moves. (Same bullshit with nickel puts too, marginally more liquidity).
20/

So that's it, that's $SKEW. It's broken- if it can be, given that it never worked. Even it's own whitepaper is lame. It was a quant dream, an intern's project, a bullet point in the minutes of a meeting that should never have been called. Please stop using it. Please.
21/