Mean Reversion trading works for equities; and currencies.
We saw in the last article how combining two simple ideas for equities produced a stable system over the last 30 years.
Can we repeat a similar analysis for currencies?
However, be warned. Currency trading is a different magnitude of difficulty to equity trading. Currency traders have had a real tough time since 2008 (take a look at the BTOP Barclay Hedge Currency Trader Index).
As always it depends which pond you fish in. Whereas equities can see slow trend grinds, or explosive surges, currencies are much more choppy. Have you ever seen a G10 currency perform a “Yahoo party like it’s 1999” dance?
In this article we’ll cover:
- Defining Mean Reversion again
- Finding the Right Pond to Fish in
- Testing Patterns for Mean Reversion
- Constructing a real simple but well performing mean-reverting portfolio
- Using the benefit of Diversification to combine it with our Equity strategy
Defining Mean Reversion
Recall: Mean Reversion Trading means fading strong moves. Usually towards their points of origin, the mean of the price series.
In this article series we covered two approaches:
- Look at the 5-day moving average (one week seems magical across assets) and trade from the other side
- Look at sequences of up and down periods.
Let’s apply these two concepts to currencies as well.
And let’s start out with EURUSD.
Why EURUSD? Well, it’s considered to be one of the most ‘technical’ currencies (at least anecdotally). More tangible characteristics: it’s certainly the most liquid, has a low/bid ask spread, and for the purpose of testing our ideas, it’s certainly exhibited strongly trending, range-bound, high-vol and low-vol environments. This makes it a good beast to try out at first.
Testing 5-day MAs on EURUSD
Recall, that the 5-day moving average approach was to trade in the direction opposite to the short-term trend. Meaning we were long if EURUSD is below its moving average and short if it is above.
You might ask, why 5-days. You can certainly vary this. However, I felt that since we fixed this period in the previous article, it’s a good example of how to look for universal properties, and not get bogged down in parameter searches.
The (minutely) data was obtained from ForexTester’s historical data service, which is sourced from a list of brokers. This is a good test to have. If different data sources provide very similar results, you know that you are not dealing with some spurious data quality issues.
The Sharpe ratio here is at 0.5.
So what about GBPUSD? Or USDJPY?
This is what it looks like for the two:
Not too good.
Now this is important: Looking for the right pond to fish in!
What do I mean by this? Forget about the majors for starters.
Would you recognize this pair without me telling you what it is?
It’s actually CAD/NOK. What’s interesting about this pair: almost no retail broker will show it.
Furthermore, both CAD and NOK as economies are strongly related due to their oil production. So it makes sense for them to strongly related.
Let’s use the 5-day MA method from before:
The Sharpe Ratio alone on this is 1.16! Pretty impressive.
So here is an exercise: find as many “off” pairs as you can think of. Obviously you will have to construct them. CAD/NOK = USD/NOK divided by USD/CAD, where you can obtain the data from sources such as your MT4 History Center.
Mean Reversion: Trading Against the Trend
We’re going to stick with the concept of 5 business days, better said a week (the signal over dailies is too noisy, and not much comes of it).
Similar to the equities setup we’re going to try something really naïve.
If the last week’s currency move was up, go short, and vice versa. If the last week’s currency move was down, go long.
At first sight this might not seem like much.
Here are the results for this approach for 28 pairs typically found on brokers (same data set as before with FXOpen as the source).
Aggregating these results we obtain:
Note that we haven’t cherry picked any of the currencies which had underperforming periods.
There might be some arguments to be made for selecting only those that visibly clearly exhibit ‘mean-reverting’ characteristics, such as the CHF pairs.
Performance Measure of this Strategy
It turns out the Sharpe Ratio for this strategy is at 0.7.
That is pretty astonishing.
And there are some key subtleties here. The most important one is that we are not trading one single currency pair.
Instead we have a currency mean-reversion index.
Similar to the equity setup where mean-reversion on a single-stock would not have been as powerful, however in aggregate the signal becomes very strong for the index itself, the S&P500.
Looking at the correlation of this strategy with our equity strategy: -5%!! Driven primarily by the US debt downgrade shock in August of 2011, where currencies shocked the other way to equities.
And this is an indication that we might want to mix it up, and put these two together.
Adjusting for volatilities, we obtain:
With a Sharpe Ratio of 1.17
No that is really not too bad!
We’ve covered mean-reversion on currencies in this article.
And as we indicated at the start, trading currencies can provide a much tougher time. However, combining them with other assets provides great diversification.
Even more importantly, none of the methods we’ve tackled so far are ‘rocket-science.’
That’s not the point of trading.
The point of trading is to find something that provides juice and systematically extract it.
See You Next Time…
We’ve covered the equity portfolio of our Consistent Trading Portfolio.
Next up will be bonds.
Bonds are much more tricky to deal with, since they are finite maturity products, that pay coupons on a regular basis. Nowadays you can get useful data from bond ETFs, such the AGG, TLT, SHY, etc.
The biggest argument levelled at these ETFs is their short history, and the fact that they’ve been trading during one of the biggest bond bull markets.
Many say that now with rising rates we’ll see the end of their profitability and they could even be drag to include in a portfolio.
So, the next part of this series will look at putting together an index which we’ll calculate off available government interest rate data going back 70 years or so. And then we’ll have some fun looking at bond behavior and its contribution to our portfolio.
So, until next time,
If you have enjoyed this post follow me on Twitter and sign-up to my Newsletter for weekly updates on trading strategies and other market insights below!