VXX is a basket of two closest VIX futures contracts. If you take a look at its performance, it goes down 99% since opening in 2009. So there is an obvious opportunity to make some money shorting VXX. Let's try a simple strategy, go short VXX all the time. A performance equity (logarighmic scale) will look like that:
It's tradable, but strong drawdown of 2011 makes holding such a position really scary. Let's try making a better strategy.
Why VXX price declines? This is happening because of contango on VIX futures. So why don't we use a value of this contango as a coefficient for our short position? Ok, I calculated for each day a price for a two-contract basket, just like one they use in VXX. The contango is logarithm of relation of this backet price to original VIX index value.
Here is a histogramm of contango values, you can see that VIX is in contango most of the time, but backwardation tail (left side of distribution) is really heavy.
Now we have a controlling factor for our short position. When we have steep contango we hold strong short, when we have a little contango, we go a little short, when we have backwardation, we go long VXX.
The equity of such a strategy:
As you can see, 2011 drawdown is completely cured.
But what if we use contango as a binary factor, i.e. go full short whenever we have contango no matter how steep it is and go full long whenever we have backwardation? Here is the picture:
Well, it looks very straight until 2013 and probably drawdown of 2013 is even smaller than drawdown of 2010 in previous picture. This result is a bit surprising, I don't know why is it better to ignore the value of contango. But seems it works.
All this strategies have research purpose, no transaction cost included. Do not use anything before studying transaction costs impact.
It's tradable, but strong drawdown of 2011 makes holding such a position really scary. Let's try making a better strategy.
Why VXX price declines? This is happening because of contango on VIX futures. So why don't we use a value of this contango as a coefficient for our short position? Ok, I calculated for each day a price for a two-contract basket, just like one they use in VXX. The contango is logarithm of relation of this backet price to original VIX index value.
Here is a histogramm of contango values, you can see that VIX is in contango most of the time, but backwardation tail (left side of distribution) is really heavy.
Now we have a controlling factor for our short position. When we have steep contango we hold strong short, when we have a little contango, we go a little short, when we have backwardation, we go long VXX.
The equity of such a strategy:
As you can see, 2011 drawdown is completely cured.
But what if we use contango as a binary factor, i.e. go full short whenever we have contango no matter how steep it is and go full long whenever we have backwardation? Here is the picture:
Well, it looks very straight until 2013 and probably drawdown of 2013 is even smaller than drawdown of 2010 in previous picture. This result is a bit surprising, I don't know why is it better to ignore the value of contango. But seems it works.
All this strategies have research purpose, no transaction cost included. Do not use anything before studying transaction costs impact.
This comment has been removed by a blog administrator.
ReplyDelete