SK10 vs 25D RRN

Discussion in 'Options' started by VolSkewTrader, Jan 28, 2019.

  1. This is just the tip of the iceberg. Most people aren't sharing their real knowledge in this industry. Everything is siloed.
     
    #11     Jan 29, 2019
  2. srinir

    srinir

    How does Skew calculation calculated from historical returns compare with the Skew calculated from options. In my opinion all these market based probabilities are not much better than simple historical calculations.

    In your other thread I posted equity skew from Minny FED. This one is the same for crude.

    Snap35.png
     
    #12     Jan 29, 2019
  3. Not sure if observed skew behavior is a good predictor of future volatility (standard deviation on your chart). From what I've seen, they usually move in tandem with each other, with demand for downside or upside (changes in the skew) often lagging the violent moves in the underlying. For example, the OTM put vols don't explode in anticipation of a market gap decline. The panic to purchase puts will happen only during or after a market meltdown. And the OTM calls will only catch a bid (IVs increase more relative to same delta put IVs) when a sustained rally has been confirmed by the market.

    Also, when futures are range bound and volatility is creeping in, the skew will behave independently of volatility
     
    #13     Jan 29, 2019
  4. Just curious, how would you calculate Skew from the historical returns of the underlying? That doesn't sound possible.
     
    #14     Jan 29, 2019
  5. srinir

    srinir

    It is just a third moment, Variance is a second moment and Kurtosis is a fourth moment
     
    #15     Jan 29, 2019
  6. Got it. You're just taking the normal distribution of historical returns of an asset and determining a Skew from observed data. Well, as you know and its been proven that implied (expected) volatility is a much better predictor of future volatility than historical. Although implied vol typically trades at a premium across all products and is often mis-priced, it's much better to look at a forward looking instrument rather than something that happened in the past to determine the expected future behavior of an underlying.

    The Skew Index is a "noisy" indicator, and so far seems to be a lousy predictor of a "fat tail" event. I don't think you can gather accurate probabilities of a Black Swan event by just measuring the steepness of far out winged puts (5 delta <), which I believe the CBOE is doing. Sub 10 readings of the VIX are probably a better predictor of a future "tail event."

    The Oil Skew Index (if it had one).I think would in general be a lagging indicator, as it prices in a "fat tail" event when it's actually occurring. There's opportunity in commodities to ride the panic wave before it peaks, and then "fade" it during market capitulation. Easier said than done. It's a game of musical chairs.
     
    Last edited: Jan 29, 2019
    #16     Jan 29, 2019
  7. What do you plan on doing with all this information?
     
    #17     Jan 29, 2019
    nooby_mcnoob likes this.
  8. sle

    sle

    I suspect that he intends to make money, but I could be mistaken.
     
    #18     Jan 29, 2019
  9. Thanks genius. I must be mistaken, tracking/plotting 25d risk reversals is a surefire way to make money. Everyone can go home how!!
     
    #19     Jan 29, 2019
  10. sle

    sle

    Well, the general idea is exactly that - 1. get information, 2. process it, 3. profit. The rest is details which nobody, myself included, wants to share. I wonder why?

    There are several other ways of looking at historical skew realization, some better than other and looking at different metrics. Here is one, albeit a bit labour intensive for linear skew in equities. First, build a model for directional dependence of realized volatility using following regression:
    Code:
    log( RV[t, T| T] / IV[t, T|t, k=ATMF] ) ~ A * log( S[T]/S[t] )
    
    Now, for any given point in time, you can compare the slope A from the historical model to the "local vol" model B which would be something like
    Code:
    log( 0.5 * IV[t, T| k=K, t] / IV[t, T| k=ATM, t] + 0.5 ) = B * log(K/A)
    
    You can think of more ideas along the same lines, I am sure.
     
    #20     Jan 29, 2019
    ErnstAugust and srinir like this.