This week we started work on a new research line: statistical arbitrage in forex.
The first tests were useful, but not because they were immediately strong. They were highly unstable. That is exactly the kind of result we want to find early, before a strategy idea is dressed up as something more reliable than it really is.
The work has now moved toward a more practical direction: a momentum-driven alpha model supported by external Python-accessed data models. The aim is to assess the relative strength of currency carry trades, then use momentum indicators as entry signals rather than treating carry as a standalone reason to trade.
In plain terms, we are asking a few simple questions:
- Which currency carry opportunities look strongest?
- Is momentum confirming or rejecting the carry idea?
- Does the signal survive once it is passed through the Penfjell risk framework?
- What happens when it is tested on learning datasets rather than judged from a single attractive example?
The current stage is intensive testing. The model has been connected to the risk framework so we can look beyond headline returns and focus on stability, drawdown behaviour, signal quality, and whether the approach has enough structure to justify deeper development.
It is early research, not a finished product. The useful point this week is that the unstable first pass did not end the work. It helped shape the next question: can carry strength, momentum confirmation, and disciplined risk controls produce a cleaner FX research candidate?
That is what we are testing next.