Re:Robinhood Traders Discovered a Glitch That Gave Them ‘Inf...
by MuffinFlavored on Tuesday November 05, 2019 @09:43PM
(#21459257)
You made a blog post that 10x leverage would be good given the past 12 months (where the market has mainly been up-up-up)? Doesn't that 100% ignore the effects of drag when the market goes down-down-down?
Re:Robinhood Traders Discovered a Glitch That Gave Them ‘Inf...
by smabie on Wednesday November 06, 2019 @03:29AM
(#21460666)
First of all, the market has not been exactly “up-up-up.” There has been massive volatility and drawdowns in the last year.
Second, my analysis was from the beginning of 2018 to present. Using that data, the ideal leverage ratio is 2.99x.
Third, the ret/var model put forth in the post assumes a log-normal distribution. The less log-normal the returns are, the less accurate it becomes. Excess kurtosis or skew will definitely affect the accuracy of the model. So the model would probably be fine if the market was up-up-up or down-down-down, I haven’t encountered any periods in which the non-normality of the distribution is so severe that it completely breaks the model.
https://smabie.github.io/posts/2019/10/04/vol.html
In short, assuming a normal distribution, your max leverage ratio should be the future expected returns divided by the expected variance.
I’m not at a computer right now so I can’t run the math on real S&P returns, but assuming 10% returns and 10% vol and a normal distribution, the highest leverage ratio you would ever be able to justify is 10/10^2 or 10x leverage.
In my blog post I looked at the beginning of 2018 up to present. The ideal leverage ratio according to my equation (I’m sure it’s not novel, but haven’t been able to find anyone else talking about it), is 2.99x.
Of course, the more skew or kurtosis the distribution has the less accurate this equation becomes.
Right now I’m working on a hypothetical S&P ETF that adjusts leverage bases on forecasted volatility and returns. Almost impossible to get right, but my preliminary model has outperformed (not risk adjusted) any fixed leverage I’ve thrown at it.