A quick post today to let you know that my base strategy from the Economic Pulse Newsletter has been tested by AllocateSmartly and is now available on the platform. I wanted to put my base TAA model to the test on an apples to apples basis against the best TAA strategies available on the best TAA backtesting platform. I felt it was an important to validate my results and test the strategy against others on an independent basis. I’m pretty pleased with the results.
I’ll let Allocate Smartly take it from here. You can read the full post on the test of SPY-COMP here. I’ll just post some of the summary results and make a few comments. First, the big picture results.
Basically, market beating absolute and risk-adjusted returns with relatively low turnover. But not only low turnover, the average trade length of the strategy is quite long so 91% of the gains from the strategy are long term capital gains or dividends.
As far as performance vs the other 55 strategies tracked by AllocateSmarlty, in terms of annual returns, over the full period of test, SPY-COMP is tied for 6th place. Over the last 20 years, SPY-COMP is tied for 5th place. And over the last 10 years, it is in second place. SPY-COMP is a 100% US centric strategy and ideally should be combined with other strategies to build a complete portfolio. In Economic Pulse, I use SPY-COMP as the base strategy and build upon it by adding other asset classes to increase diversification and performance. You can also do that with other strategies on AllocateSmartly.
That’s about it for today.
6 Comments
john · April 25, 2020 at 10:47 am
Hi Paul, I see your strategy on Allocate Smartly is based on Monthly changes
however, they also provide daily updated information. What are the backtested results had you traded the signals daily, or even weekly?
I’d be very interested in following this weekly and wonder if the “whipsaw” would distort the impressive results
paul.novell@gmail.com · April 27, 2020 at 12:10 am
Not as good as the monthly in general. The weekly does have lower drawdowns but also lower returns (about 1-1.5%). And more whipsaws of course.
Paul
Todd · April 28, 2020 at 12:35 pm
Hi Paul,
In a previous post on your site there was a backtest of SPY Comp from 1973 through 2019 (essentially the same as the new AS efforts). AS is arriving at annualized return of 13.6 and your post had a 16.20% CAGR with significantly different sharpe ratios.
Do you know why there was such a large difference which such a small time difference in the test?
paul.novell@gmail.com · April 30, 2020 at 3:18 am
Hey Todd,
Well, a few reasons. My backtest were not nearly as robust as AS; data sources, slippage, etc. That’s one of the primary reasons I wanted to subject the model to such a stringent test. Also, I must have fat fingered something in a spreadsheet somewhere, that number is too high. There are some differences in the implementation but not enough to make that big a difference. The numbers I have currently published for SPY-COMP, see here – scroll down to the long term performance section, for the 1973 to 2019 time period are 13.39%.
I’ll go back and find and fix the old errors. Thanks.
Paul
Brandon · May 3, 2020 at 4:31 pm
Hi Paul,
Are you concerned that recessionary periods are becoming shorter and shorter in the modern age of algorithmic trading and the lightning speed of information transfer? With some exceptions, of course, it seems recessionary periods are getting shorter and shorter.
The Great Depression: 4 years
Dot Com Crash: 3 years
Great Recession: 2 years
COVID Recession: Unknown, but already maturing at record speed, from top to bottom and back up again
Given how fast information travels now, it also makes intuitive sense that crashes and recoveries today would happen at quicker speeds.
If this assumption is true, does this limit the effectiveness of trend following systems, in your view? Early data is already showing trend following systems largely underperforming the 60/40 benchmark during this COVID-induced drop. This is unprecedented for a drop this large, and its due to the neckbreaking speed at which everything has transpired.
Would love to hear your thoughts.
Thanks!
paul.novell@gmail.com · June 25, 2020 at 12:20 am
Hey Brandon,
I just found your comment in spam. Sorry about that and the long delay……
Basically, yes that is a concern for trend following approach, that the models continue to get whipsawed, like they have since 2009. But what is interesting
is that in every event in your list many TAA models have outperformed in every one. All models are not created equal. Even in this ultra quick COVID recession, about 60% of the TAA models I track
are outperforming the SP500, and even more are outperforming a more global benchmark (like GAA). About 32% are outperforming the 60/40 portfolio. This makes the case for good diversification
among models with good characteristics.
Also, I would caution on calling this current cycle over. It can’t be called over until the broad market, like the SP500, makes new all-time monthly highs.
After this then we can truly measure TAA performance during the COVID recession.
Finally, it is not the recessionary periods that are the real issue, especially as you noted, since they are becoming less frequent. It is the false signals outside of recessionary periods that are the biggest detractor on TAA performance. So, it is important, in my opinion, to try and weed out the false signals from the real ones to avoid the whipsaws. That is what I try and do in my models.
Paul
Comments are closed.