It takes time to research financial markets. From a quant point of view, the aim is always to build models which are generalised, which can work for many assets, whether they are indicators, trading strategies etc. One such model is a trend following strategy. The same trend following algorithm can be applied to futures across different asset classes (and beyond that). In practice, there are going to be complexities, such as the differing levels of liquidity of each futures contract, which means that we might have to tweak things, but by and large the algorithm would be pretty similar.
Does this mean that we should always try to build generalised models that can work across many asset classes? In an ideal world yes! After all, it means that such a generalised strategy is likely to be more scalable, which is especially important for larger funds. Even if we are looking at a specific asset class like FX in isolation, the question is whether we should model all the currencies together, or models for a specific currency (or groups of them). In practice, if we are always chasing more generalised models when doing financial analysis, we are likely to be missing a lot.
In FX, there might be very fundamental reasons why a certain model is appropriate for some currencies and not others. Take for example trying to model FX based on shifts in interest rate differentials which are a proxy for monetary policy changes. In developed markets this has historically been profitable, because these currencies appreciate when yields go higher. If the same approach is applied to emerging markets, it is not such a great idea, because during times of risk aversion higher yields can be observed because investors are dumping the local currency and local bonds at the same time! Does this mean we should ignore interest rate differential shifts to model FX? No, it just means we need to be aware that the approach doesn’t work for all currencies.
We could also drill down further into looking at models for very specific currencies. Here again there might be very good reasons why a certain approach could be useful for some more than others. Take for example currencies of countries which are very large commodity exporters, such as AUD and BRL. It makes intuitive sense that commodities prices more broadly should impact them. However, it doesn’t mean that the same approach for other countries is going to be as successful.
In a sense, we should try to aim to have more generalised models for trading FX and also more specific ones too. The flipside is that researching more models costs more in terms of time. There also might be issues in terms of capacity constraints for models which can’t trade a lot of assets.