Tail risk as a trading edge

Discussion in 'Trading' started by bashatrader, Jun 5, 2019.

  1. I always get to know about funds with high performance only after the fact. :(

    An example is AQO managed by Nigol Koulajian and his team.

    Is it possible to win this big with tail risk trading as an edge?

    Is the method used by the author good enough to determine when the edge is gone?

    Regardless, a very interesting article.
     
    tommcginnis likes this.
  2. tommcginnis

    tommcginnis

    Their equity curve looks like their biasing long on equities and then long on vol in market turns. Which means not so much in counter-trend or even short equity positions...
    Just eyeballing it, but I seems not-so-much centered on tail risk. No hint (to my eyeballs)...
     
  3. R1234

    R1234

    The verbiage about 'volatility expansions' and trading 'tail risk' appears to be mostly marketing bullshit.

    The track record is highly correlated (~0.7) to simplistic trend following.

    upload_2019-6-5_12-36-33.png
     
    Last edited: Jun 5, 2019
    bone and tommcginnis like this.
  4. gaussian

    gaussian

    When you get rid of the marketing bullshit 'capitalizing on volatility expansion' and 'tail risk as an edge' becomes 'buying calls and puts'.

    Additionally, the use of statistics in this article is suspect:

    They don't discuss their bootstrapping method. Time series (including the return series), especially considering they didn't explicitly specify logarithmic returns, have small amounts of internal autocorrelation that, when resampled improperly, become magnified and can ruin the result. If I recall correctly from my computational finance course, the only reason the bootstrap resampling method they used work is it relies exclusively on the assumption the logarithmic return series comes from the normal distribution. Again, they do not specify this, and so their test is highly suspect. This could almost qualify as p-hacking which would make this a shill article.

    There is a method for bootstrapping time series called block bootstrapping. It doesn't appear to me they used this.
     
    bone and tommcginnis like this.
  5. ronblack

    ronblack

    I found the article interesting. If you read it again you will notice the following comments by the author:

    "We can never be 100% sure and also the test reliability can be questioned."
    "We conclude that for the period 2009-2019, we cannot reject the null hypothesis."

    As it appears then this cannot be a shill article for the fund. Your statement is silly based on the article conclusions. If it were shill article, it would stop at p-value = 0.006.

    I agree about problems with bootstrapping but according to claims of fund this is not trend-following but short-term volatility expansion trading with average trade holding of few days and auto-correlation in monthly returns can be ruled out.

    I disagree with the article that this is an edge. It is more of speculative trading where something can go very wrong at any time.
     
    tommcginnis likes this.
  6. tommcginnis

    tommcginnis

    I'm going to reiterate my vote that the fund *looks*like* long-only in equities and then long-only in vol., but "could go very wrong" part -- I'm going to disagree with: we've had non-market shocks through the last few years, and the fund's path shows hit AND recovery -- it's not just long-only "everyone's a genius in a bull market" effect, nor the vol-oriented reverse of that. I'd love to test out an either-or regime of long-equities/long-vol, but it would just have to go on a sticky-note for execution "sometime"...