The use of next-generation technologies in foreign exchange (FX) algo (algorithm) trading has been a hot topic of late, although many of the practical implications still tend to be misunderstood. FactSet’s Christopher Matsko, Head of FX Trading Services, recently spoke to FX AlgoNews about trends in the FX industry and available trading solutions.
Q: Do you see AI and ML as key buy-side tools today?
Despite all the buzz, in reality, we are not yet seeing artificial intelligence (AI) and machine learning (ML) being used to nearly the same degree as all the hype would suggest, particularly on the buy-side. Asset managers are still adopting FX algos in various capacities, whether in basic forms of simple WMR [WM/Reuters benchmark rates] algo executions or picking up more complex broker or proprietary strategies for their executions. These tools, while no doubt quite clever, fall short of true AI/ML, and perhaps it’s important to draw that distinction somewhat as the lines are sometimes blurred.
In that sense, it’s important to qualify the difference between real machine learning via real neural networks versus what tools such as smart order routers are being used for today. The latter is being fully embraced by the market via automations centered around the integration of post-trade analytics into a pre-trade decision process. The process in and of itself might look like machine learning, but it’s just sophisticated automation.
Q: Are intelligent algo wheels gaining momentum in the FX industry?
We recently hosted a roundtable with more than a dozen head traders where algo wheel usage was a highly debated topic. We discovered that the first school of thought on this area was that even though a firm might have curated their wheel in a very particular fashion, the trader still doesn’t want to let the wheel make the decision for them wholesale. The same trader is in possession of high-quality transaction cost analysis (TCA) metrics that should help them make the appropriate determinations around optimal execution.
The second school of thought centers on the notion that algo wheels, when used judiciously, can effectively reduce the potentially costly effects of trader bias. This argument, among our panel, seems gradually to be gaining acceptance. We believe that with the advent of more accessible and reliable data, and the systems to consume it and present it meaningfully at point-of-trade, the usage of FX algo wheels are, if anything, likely poised to increase.
Q: What are FactSet’s plans for introducing AI/ML to its FX product suite?
FactSet has a well-established cognitive computing team that is leveraging AI/ML/NLP [natural language processing] to create sophisticated solutions across its client base. There are also third parties we are talking to that have established solutions in this space. It is no exaggeration to say that we see great potential in leveraging predictive analytics for the enhancement of pre-trade decision-making capabilities—both manual and automated—but we also want to make sure that there is a level of transparency in the decision-making process that provides clients a level comfort to use these technologies with confidence.
Q: Can clients leverage their own AI infrastructure to create and deploy algos via FactSet?
From the beginning, we have wholly supported our clients’ ambitions to combine vendor technology with proprietary (in-house) expertise to achieve something not possible without each respective element working in concert. Needless to say, if clients have access to their own AI-based tools and they plan to integrate those into our execution platform, then we support it fully and robustly.
Beyond the APIs themselves, we provide our clients with tools to test their strategies before implementing them, and furthermore, we allow them to augment FactSet proprietary algos to fit their particular trading needs.
One of FactSet’s core tenets is supporting innovation in technology, which includes supporting proprietary client initiatives. For us to do that effectively we need to stay ahead of the technology curve. It is by no means a static process. Nor is it a mere tagline. It requires considerable investment and re-evaluation to keep our APIs at the forefront of what the market demands.