I think there is a very interesting topic that has not been getting much attention even on a long history forum like Babypips.com.
I wanted to start a thread where a discussion starts about asset allocation.
This issue is still valid if someone only trades spot FX, but of course it gets more interesting if someone has Commodities, Stocks etc. in his/her portfolio.
I would like to discuss who has which assets in his/her portfolio and what is the percentile distribution of the different asset classes. We can share studies, articles, videos etc. and benefit from the shared work. Of course the main point is to find the right combination to find maximum reward and acceptable risk. And as we know this combination can differ for everyone based on their level of risk acceptance.
From my perspective, a big factor of asset allocation when it comes to FX is correlation.
An example is if you short CAD/JPY and AUD/JPY simultaneously, you are more or less just doubling your risk since the two FX pairs have a correlation of ~0.921. On the other hand, long trades in EUR/AUD and AUD/CHF at the same time essentially hedges each other (-0.967 correlation coefficient). Now, apply that to ~30 FX pairs, that all share a certain degree of correlation; a total of 870 (30^2 - 30) correlation relationships.
Another scenario that needs to be considered is the correlation between trading strategies. As a systems trader, let’s say you have two trend following strategies that are based on different logic. But upon examining the equity curve of the two (aligned by date), you may find that they have a correlation > 0.85. Again, many traders will think they are diversifying their trading when in reality they are doing the exact opposite.
I am interested to hear what you, and others, think about this. Do you see correlation as an issue in risk/portfolio allocation? And if so, what is a realistic solution?
For the example I take your AUD/CHF and EUR/AUD trades. For me this is only hedging without a chance of winning. If a correlation is 96.7% then for me it means that it is almost impossible to win. If we even take the spread there and the carry trade cost (which is negative because we pay more on EUR/AUD than what we recieve on AUD/CHF) then it is most likely an automatic loss.
When we talk about asset allocation then as you said, one point is to reduce risk with a hedge built into the strategy of combining many assets, but it is possible to come out as a winner. In the case above it is very tough to come out even at breakeven. I would either go long EUR/AUD or AUD/CHF, but not both at the same time. Actually I that particular case I would go only long on EUR/CHF to save costs.
Of course you are right, it is hard to come out as a winner when hedging. So maybe it is better to use the expression “natural hedging” when trading about multiple asset clases. The example we used above is clear, it is hard to come out as a winner in that case.
But if you trade stocks, bonds, currencies etc. at the same time then there should be a built in natural hedging in the strategy and when we make a win with one asset, the goal is to have less loss with the other asset classes so we come out ahead in the end.
My opinion is before we concern ourselves with multiple markets, let’s just focus on the one in the beginning. Basically what I am asking is if you have some kind of proposal to solve these FX-related problems (FX pairs and strategy diversification). Once we have a solution for the asset allocation of the FX market, then we can look ahead onto inter-market asset allocation.
yes, there is a possible solution. Lets say we get a long signal for AUD/CHF and a short signal for EUR/AUD. You already mentioned the very high correlation between the two pairs. So, instead of taking both signals, we can take AUD/CHF long and lets say we can take EUR/NZD short. The two trades might still look a bit similar and surely there is correlation, but I do believe some natural hedging is already built in into this very simple portfolio. Maybe the EUR/AUD signal was better than EUR/NZD, but we rather give up a little from the possible higher win for reducing the risk with taking two almost identical signals.
I think you are missing the bigger picture. The first two examples I had is just a very small piece of the equation. I’ll re-state my original concern when it came to asset allocation:
Let’s say as a new trader, you want to limit yourself to risking 0.5% of your balance per trade. You want to trade just 5 FX pairs, but acknowledge the relationships (correlation) that exist between the exchange rates. So, you choose to pre-select the 5 FX pairs prior to trading, as part of your trading plan. How would you do that, while ensuring your holdings are as diversified as possible?
In an example scenario, perhaps you already have some confidence with the EUR/USD, and EUR/JPY, and they are only 20% correlated, so you add these two to your trading plan. Next, you find that the USD/CHF is only -7.9% to the EUR/JPY, so you consider adding it to your portfolio - but wait a second, the correlation between USD/CHF and EUR/USD is -95.1%! So now, with these three FX pairs in your trading plan, there are some conflicting relationships.
Imagine how these relationships would look if you had 4, 5, 6…30 FX pairs. This is what I’m trying to get at. 1 or 2 trades is negligible in the bigger picture, I’m concerned about diversification in FX in the long run. How would you solve this? Is there any theories that can help (MPT comes to mind)? Also, is keeping 0.5% risk per trade actually a good idea?
Again, this is just the first scenario with FX pairs themselves. There’s still the concept of correlation among trading strategies to consider afterwards.
One question I’ve struggled with lately in developing my plan is determining the appropriate # of periods for the correlation. Given correlations can change from their historical vs. 1 year. vs. 1 week, etc.
What # of periods or timeframe do you consider when trying to look at correlations? Right now, I have the ones I listed above but have no idea which ones to consider when looking to place a trade with non-correlated pairs.
Speaking as a systems/algorithmic trader, the number of periods to calculate correlation should be dependent on the trading model that requires it. I prefer to have my models find its own optimal periods to measure correlation. There’s an endless amount of ways to do that; as long as it is logical in your trading plan.
For example, if you’re using a 200-period moving average in your strategy, then it might make sense to calculate the correlation with 200 periods (maybe more), since the value of the current bar’s 200MA data point will be based on the previous 200 bars (including itself). A fixed period works just as well (50, 100, 500 bars, etc), again, as long as it makes sense in the overall logic of your trading plan. Note: A larger period is going to require more processing power and will take more time, so if you are needing to calculate it per bar, it will slow down your code (parallel computing would relieve this).
As for a timeframe, again it would depend on the strategy itself. As you probably know, correlation can change significantly across time frames. The current correlation between EUR/USD and GBP/USD is 10.7% on the weekly, and 72.1% on the 5-minute. In my examples above, I was using the correlation coefficients for 1-day bars and the most recent 200 bars. Perhaps your strategy is deployed on the 1-hour timeframe, and it might make sense to calculate your correlation using 1-hour bars as well.
As for managing an overall portfolio, it can get a bit more challenging. Which is why I started off with this issue (took me over a year to figure it out myself). Imagine trying to minimize the correlation within your portfolio when you trade ~30 FX pairs, 8 timeframes, and 20-30 strategies at any given time. :30:
Well of course everything is relative. From your perspective the 30 pairs is the big picture and from my perspective the complete portfolio allocation. Anyway, I am not here to argue about who has what kind of picture in view, rather to find solutions and discussions:-)
If you purely want to trade the forex pairs, there is an old and long thread about it, I can highly suggest it to you. Clint made some great posts in it. The thread is however not active anymore.
Further answering your post, of course it is a lot more difficult to diversify a portfolio when we only talk about the fx market. In the COT thread Mike was making a difference between the three commodity currencies (NZD, CAD and AUD) and the rest. This can be a good starting point when talking about correlation and diversification for currency pairs.
I’m not here to argue, I was trying to engage in the discussion that you started and I wanted to ask if you had any opinions for some of the problems that we faced as traders. If you don’t, then that’s fine, but no need to get defensive.
This isn’t about whether I’m solely interested in Forex or not; the concerns I’ve brought up is market-agnostic. If one cannot even solve it in one market, how can they solve it across many? There is inter-market correlation just like correlation between FX pairs. The examples I used could have been just as easily be replaced by assets spread across the ETFs, futures, options market.
Modern Portfolio Theory is probably the most commonly used model when it comes to asset allocation within a portfolio (across all markets). The main factors are risk (measured through the asset’s volatility), expected return and covariance/correlation. It emphasizes a diversified portfolio (minimizes covariance) in order to lower the variance of the portfolio by less than the variance of any individual holding.
Candidly, every response you have provided so far has been a “non-answer”, only looking at single trades rather than from a scope of the portfolio as a whole. You talk about wanting to discuss how to balance our portfolio, but the only contributions you have made so far has been from a perspective of a few trades. The examples were simply a way to illustrate the point.
you are right about what you said. I tell you honostly, I am waiting for more participants to maybe share the work ahead of us. In another thread it has worked in a great way, producing good results. It is however hard to do it all if we are two person.
Thanks for the contribution, I will think about it how we have to handle it if we are two people. Maybe what we can start already is maybe a list of tasks which have to be analysed and solved.
I also have to say, unfortunately I am very busy lately, I started the thread already to get some people, see some ideas and opinions. You are the only one until now who gave thoughts and I respect that. We have time, this is a large amount of work, I am not in a hurry.
I even want to finish some work in another thread, but did not have the time until now.
I´ll do my best to contribute more, might take some time unfortunately.
Maybe I have a question so you can start the discussion. You said you trade 30 pairs (maybe you wanted to say 28 as that is the combination of the 8 main currencies). Questions: are you always in the market with all pairs? What is the maximum number of trades you do? How much do you risk per trade? If you do have a max. number of trades, lets say 7, do you have a rules how many times are you allowed to you one single currency, like the USD? I am interested for the answers to get to know a bit your trading style.
I said ~30 FX pairs as an example. But yes, you are correct, 28 pairs exactly if we are talking about purely currency pairs.
No, I am not always in the market with all pairs, but there is always the possibility.
Do you mean maximum number of trades at once? I’m not sure. But if you’re talking about number of trades in a given period then 8000-10,000 trades per month.
My position sizing is adaptive depending on the trading models, correlation, and risk (among other factors) at the current time. Percentage-wise, anywhere between 0.01% and 0.2%.
No fixed rules for maximum number of simultaneous trades in a given currency. Again, depends on my portfolio at the time. Generally speaking, it’s diversified enough that I do not need to worry as I am confident in my diversification strategy.
I am an algorithmic/automated trader; trading is 99.9% automated. I have a collection of a few hundred trading strategies, some profitable, some not. Based on market conditions (news, correlation, regime, volatility, risk, strategy performance, etc), I will run 20-30 of these strategies at once across a range of FX pairs. Risk and so forth is balanced real-time based on a set of rules given. Everything I said above, is just a generalization and is true for 90% of the time.
If you look closely, right where all colours enjoin on the first low, then see the blue line and imagine that it is an indicator of risk right now, it is reflecting sentiment this very moment.