Q&A on Volatility Curve Model

Recently, I have received a bunch of questions on the volatility curve model that I introduced a few months back. In this post I wanted to answer a few key questions on the model.

Current Status

What has the model done during this most recent period?

The details are for subscribers only but the model did trigger risk off and all of the options in the model are up for the year and outperforming. The model has performed as you would expect during a time like this.

Trading

Why does the model trade at the closing price of the following day from the signal? 

For two main reasons. It’s the easiest way to implement the model and the performance is about the same, actually slightly better, than trading at the close of the same day of the signal. Of the 4 options I modeled, trading at the next day’s close is the best option (for CAGR and DD), followed by the previous close, followed by the mid-point price of the following day, and lastly trading at the next day’s open. The first three options are ok. Whatever works for you. Don’t trade at the open. See table below. Any of the 4 options are still way better than buy and hold.

The reason the next day’s close works the best is that there is a short term mean reversion effect after extremely volatile moves. Over more recent periods, since 2012 where markets have been calmer than the 2008 to 2011 period the mean reversion effect has diminished and the previous close has performed about the same as the next day’s close.

Trading during the next day is easier than the previous close because you can wait for a clear signal from the model, instead of getting the signal before the close in order to get your trades in. Sometimes, the signal can be quite close to a trigger until the last minutes of trading. The additional advantage of trading at the next day’s close is that you can use MOC orders which makes life way easier as well.

How do the total trades breakout by year?

I listed all the individual trades in my previous post but here is a breakdown of the number of trades per year for the model. All versions of the model have the same number of trades since they’re based on the same signal. Over the last 2 years, since the beginning of 2018, as markets became more volatile, the model has trade more.

Performance Details (performance stats below through Feb 22, 2020)

What is the average drawdown of the market before the signal triggers?

It varies by what assets/ETFs are used. I calculated the drawdown of every risk-off trade and took the average. See the table after the next question below. In general, markets appear to be reacting faster over the last few years and the model’s average drawdown before a risk-off trigger has gone down in line with this.

How does performance differ between volatile market periods, like 2008 to 2011, and more calm periods?

I broke out model performance into two periods, 2008 to 2011 and post 2012. See table below for details. The model outperforms more during volatile periods but still manages to outperform during more bullish periods as well.

Is it possible to combine the volatility curve model with the SPY-COMP model?

Yes, it is and it works pretty well. See table below for the results with just one of the asset class combinations. The combination improves returns, drawdowns, and the average drawdown to risk-off. In general, during recessionary or economically weak periods you stick with SPY-COMP otherwise you use the Vol Curve model.

That’s about it. Those are the main questions I wanted to answer. If you have more questions leave a comment or drop me n email. If you’re interested in the model it is available as part of the Quant Pulse service. And more information on SPY-COMP see the Economic Pulse Membership.