A long-time member who has been a valuable source of feedback over the years sent us the following note about the most recent strategy added to the platform: Gold Cross-Asset Momentum.
The strategy has performed poorly relative to other strategies on the platform. You’ve turned down other stuff that was marginal like this, so I’m surprised it made the cut.
He’s right. Viewed in isolation, performance has been inferior. This is a good opportunity to talk about how our thinking has evolved on deciding what strategies to add to the platform.
Core TAA is saturated, so the hurdle is higher:
We track a lot of strategies that we would characterize as “core TAA”:
Start with a bunch of asset classes. Apply some flavor of trend-following/momentum and some weighting scheme to allocate among assets, and voila. There are classics like Faber’s GTAA 5 and Alpha Architect’s RAA, as well as newer choices like Hybrid Asset Allocation.
When we think of TAA, our first thought is this class of strategies. They differ in small ways and big, but they all leverage this same core idea.
For us to add to this list of strategies, the hurdle is higher. Members benefit little from adding another core strategy unless historical performance has been strong (and robust, more on this later).
A lower hurdle for novel strategies:
The more novel a strategy is, the lower we’re willing to set our hurdle. There are two reasons for that, one subjective and one quantitative.
Subjectively, it’s just more interesting for members to cast a wide net, tracking many different approaches, and understanding what is and isn’t working now.
Quantitatively, there’s value in combining dissimilar things. Just like investors should diversify across assets, they should diversify across strategies as well. This strategy zigs when that strategy zags, and presto, portfolio risk goes down and risk-adjusted return (ex. Sharpe Ratio) goes up.
“Average pairwise correlation”, or a strategy’s average correlation to all other strategies we track, is a simple measure of how novel a strategy is. All other things held equal, adding strategies with lower correlation to other strategies in a portfolio will provide better diversification.
At the bottom of this article we’ve shown the average pairwise correlation for all 100+ strategies we track (see the data). The five strategies with the highest average correlation (Metas excluded) are:
| Strategy | Avg. Correlation |
|---|---|
| Faber’s Trinity Portfolio Lite | 69.3% |
| Efficiente Index | 67.3% |
| Faber’s Global Tactical Asset Alloc. – Agg. 6 | 66.1% |
| Faber’s Global Tactical Asset Alloc. 13 | 65.9% |
| Movement Capital’s Composite Strategy | 64.0% |
We would characterize all of these as “core TAA” strategies. As a rule, we believe investors should always diversify across multiple strategies, but if we were forced to invest in just a single strategy, these would be reasonable choices.
Conversely, the five strategies with the lowest average correlation are:
| Strategy | Avg. Correlation |
|---|---|
| Glenn’s Quint Switching Filtered [Dynamic Bond] | 36.0% |
| Sell in May/Halloween Indicator | 35.9% |
| Predicting US Treasury Returns | 31.3% |
| Piard’s Annual Seasonality | 29.6% |
| Gold Cross-Asset Momentum | 20.0% |
These are not “core TAA” strategies. We are not saying they are good or bad, but they are different, and thus, they may have value as diversifiers. At the very bottom of the list is the latest strategy added, Gold Cross-Asset Momentum. And that’s essentially why it was added, despite the marginal performance.
Lastly, note that there are other factors beyond correlation that might lead us to consider a strategy “novel”, like being especially tax efficient.
Stricter about basic robustness:
On a tangentially-related note…
In the past we’ve modelled a handful of strategies that we either identified as having some structural flaw (example) or concluded in a general sense were likely overfit to history (example).
We’ve often still added those strategies to the platform, assuming that our pessimistic analysis was sufficient. Our philosophy was that seeing everything, even questionable things, was better than being limited to a curated list.
In hindsight, that was the wrong approach. We track a lot of strategies, and it’s not reasonable to require new members to read through every long writeup when designing their portfolios. In the future, we plan to still discuss these strategies on our blog but either not add them to the platform or only add them after correcting for structural flaws.
Lesson learned. As a rule, we do not remove strategies from the platform, but we will take this stricter approach moving forward.
Impact on the Portfolio Optimizer and Meta Strategies:
One last consideration…
When we add strategies with low average correlation to the platform, those strategies are more likely to find their way into the optimizations that feed the Portfolio Optimizer and Meta Strategies. It’s an inherent part of portfolio optimization for all of the reasons previously discussed: when we combine dissimilar things, it lowers risk and improves risk-adjusted returns.
Should we account for this fact when adding these types of strategies to the platform? For example, should we consider how adding a strategy will impact the historical performance of Meta Strategies?
The short answer is no. As long as the strategy has value to someone (that’s the key part) we just do the work and let the chips fall where they may. To do otherwise is a recipe for overfitting. Further, 10 years from now, how many useful strategies would we have rejected because they would have negatively affected such-and-such optimization at that moment.
New here?
We invite you to become a member for about a $1 a day, or take our platform for a test drive with a free membership. Put the industry’s best tactical asset allocation strategies to the test, combine them into your own custom portfolio, and follow them in real-time. Learn more about what we do.