A Case for Adaptive Strategies: Trading the US Dollar with the S&P 500
This post is about two things: (1) a trading strategy to go with our recent research into the connection between the US dollar and stocks, and (2) a very clear example of why adaptive trading strategies are so powerful.
Previously I showed that the US dollar (USDX) strongly influences same-day returns on the S&P 500, but this post is about future returns, and for future returns, the relationship is flipped – the S&P 500 strongly influences the USDX.
The graph above shows the USDX (blue) and the following strategy (green) for trading the USDX from 1995 to 2003: go long the USDX at today’s close when the S&P 500 closes below its 2-day simple moving average (SMA) today and short at today’s close if it closes above. This is a proof-of-concept, so I’m assuming frictionless trading (no transaction costs, tracking error with the index, etc.)
That’s a very consistent result: annual returns of 16.8% (vs -0.3% for the index) and a peak drawdown of -9.5% (vs -28.4%). But look at what happened after 2003:
The entire relationship turned on its head. The worse part is that this change was NOT related to the decline in the dollar – the decline began long before the strategy stopped working. This was a fundamental shift in the dollar/stock relationship.
Now, a trader could just flip the rules of the previous strategy, and I think that would be okay, but this post raises a larger question: what are traders to do if they are constantly threatened by fundamental changes in everything they know to be true? One solution is an adaptive strategy.
A Case for Adaptive Systems
An adaptive system (like our YK Strategy) isn’t programmed with static rules like a traditional strategy; rather, it’s designed to “learn” from the markets. In theory, as market fundamentals change, the strategy should change with it.
Just for giggles I took this strategy and put it into our YK model. I told it to only use the 2-day SMA discussed above to make its decisions and nothing else. The result…
The blue line is the original strategy result and the red line the adaptive strategy result. Note how the adaptive strategy stumbled a bit as fundamentals shifted, but eventually it adapted and continued its upward march…a clear example of an adaptive system trumping a static one.
Happy Trading,
ms
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Evolving Markets, Trading Strategies | 9 Comments
Mike,
Nice thoughtful insights. I like it. I was reading bzbtrader’s post on curve fitting today and it got me thinking about stuff and then I ran into your blog via Quantifiable Edges. How do you know when to switch or strategies or modify them? It’s easy for me to look back and figure what’s wrong and then develop a strategy based on what I should have done. However, it a whole different thing to recognize these issues and execute against them in real time. Any thoughts would be much appreciated.
LP
LP – very good question. At some point I plan on doing a series of “how to” posts on building adaptive systems. It’s a little hard because on one hand you want to give folks useful information, but on the other hand, you don’t want to give away any secrets of your own “secret sauce”. One thing I can say for sure is that I don’t think a good adaptive system is “either/or”. Meaning, in the example in this post, I don’t think the strategy would suddently switch from interpreting the rules 100% one way, to 100% the other. Rather, I think a good adaptive strategy should recognize “shades of grey” and slowly migrate from one set of rules to another…that way it is less likely to get head faked by a temporary “change” in the markets. I hope that made sense and stay tuned for (at some point) some additional posts on this subject.
Thanks,
ms
What do you mean for adaptive system? It’s clear in theory, but not in practice. Give us an example, like which are the rules of an adaptive system and so on…..
RE: to Pete – see my comments above to LP. At some point I’ll do a very general “how to” series on adaptive systems. Expect something in the next month – have to clear some another goodies off the plate. More to follow…
Happy Trading,
ms
Michael,
Are there any “how-to” resources you could point us to while we wait (e.g. trade publication articles, academic articles, etc.)?
Best,
Josh
RE: to Josh
Great question, bad answer…”no”. I’ve read a lot of thoughts on adaptive systems and while they talk conceptually about why adaptive systems are good (I think that’s pretty self-explanatory) I’ve never found one that lays out how to build one in enough depth that I would recommend it (but I’m very open to being proven wrong. Everything Mrs. Kuo and myself built was from the ground floor up based on our own thoughts.
Sorry I don’t have a better answer.
Happy Trading,
ms
@MS: I agree with you in that you have to constantly wait for the transitory phase to pass before you switch. This will prevent you from trading noise. My systems are fairly simple with very few variables. I do the whole back test/walk forward to reduce curve fitting issues. I also define trends and only execute signals based on those trends beyond the transitory phase. This is a very simple “adaptive” style system. Though I will admit I cannot qualify this as an adaptive system.
I will be interested in your posts on how to build adaptive systems. I’m not looking for your secrets but rather a thought process on building a system capable of recognizing changing market environments.
Great analysis. Thanks for the post.