Notes from TradeTech FX USA

Usually, I don’t particularly look forward to long haul flights, being cooped up in a seat for hours on end, “enjoying” plane food etc. However, this time was different, earlier this month, I was excited about going on a long flight to the USA. After all, it’s only been a few months since pandemic travel restrictions were lifted to the USA. (And yes, for regular readers of my blog, I did have some burgers during my visit!)


I was travelling to the USA, to speak at the TradeTech FX USA conference, which was being held in person for the first time since 2020, that brings together the sell side, buy side and vendors. It was the first time I was meeting many of my FX colleagues in person since 2020. A lot of course has happened since then and in this article, I’m going to write about a few of my takeaways from the event. Below I summarise a few of the discussion points at the event. Given some of the discussions were under Chatham House rules, I won’t directly quote the participants.


Trends shaping FX trading

Given the pandemic, it’s perhaps not surprising the topic of hybrid working came up for trading desks. Working from home tends to mean less screen real estate and also generally fewer real life interactions with counterparties. In previous years, crypto was somewhat niche at the event. The use of analytics to trade, in particular TCA, including pre-trade came up in conversation. Whilst, crypto is still relatively small as a market, when viewed in market cap terms, it did take a more prominent part of the discussions at TradeTech FX than in previous years. One attraction from a market making perspective is that the spreads are considerably larger than FX. Another key point mentioned, is that offering liquidity and trading it, doesn’t necessarily need to be done from the perspective of being a “crypto believer”. 


Emerging markets and USD outlook

I participated in a panel on EM, one consensus view was on Brazil appreciation. I also commented on Brazilian inflation forecasts produced by Turnleaf Analytics. I have recently co-founded a firm Turnleaf Analytics alongside Alexander Denev, my fellow co-author of the Book of Alternative Data, which is focusing on generating EM inflation forecasts, using a machine learning approached fused both with traditional and alternative datasets (contact me, if you’d like to know more). Another panel discussed the outlook on the USD, the view was somewhat mixed, suggesting EURUSD higher (and AUDUSD higher, albeit later), but at the same time, seeing USDCNH also higher. Note, that the views were expressed before the Russian invasion of Ukraine, and as a result, there was a lot less discussion about RUB.


The use of data

The theme of data and AI came up on several panels. Data can be used to solve problems. But first you need to identify what you are trying to do. Is it for example, TCA, to compare between your liquidity providers? You also need to collect pricing data first of all. Data can be analysed internally by your quants and also can be shared with third parties. When you collect streaming pricing it’s “free”. Of course there is a cost to storing it and managing large amounts of tick data. However, as noted in several panels some historical datasets can be very expensive. If there’s a popular contract on a certain exchange, the price will be governed by that. Of course you can use vendors to help with analysis, to avoid having to manage large datasets yourself, which can be challenging for smaller buy side, in particular, for example in TCA space, using a firm like Tradefeedr. The cloud also makes it easier to scale your own computation and storage. From personal experience though, you still need some level of management here with the cloud. Admittedly, the process is different to managing a real hardware stack, and depends on whether you are accessing the cloud directly or simply logging into an app hosted on the cloud, where the job is somewhat easier. It was also noted that in some cases, you simply do not have enough data to use some data analysis techniques (for example deep learning). One thing I always say to my students, is always try to use the simplest model possible, which can capture the key parts of a relationships. Using a deep neural network, when a linear regression can do a specific job really isn’t necessary.


Impact of increased electronification, market relationships and multi asset

Over recent years one trend that has accelerated has been the electronification of the FX market, in recent years, in a way which was unthinkable at the start of my FX career nearly 15 years ago. The number of spot traders on both sell side and also buy side desks has shrunk. More can be done with less manual work and there’s a trend to multi asset trading desks. The whole process was becoming a lot more data driven, with data science teams helping to crunch market data. It’s now possible to get streaming prices in large sizes such as 500mm USD, which would have required manual RFQs many years ago. There was also a move to a transfer of execution risk to the buy side, in terms of execution using algos. The various polls taken at the event, however, suggested that there was still a way to go in terms of usage of algos, in particular in areas such as NDFs.


Collaboration between buy side, sell side and vendors

The buy side today is faced with many choices, when it comes to technologies. The panel discussed the question of build vs. buy when it came to technologies. One thing that has been difficult in the past, is that vendor systems tend to be more closed, which made it difficult for technologies to communicate with one another. From many of the vendors on the panel, they stressed how they sought to make their platforms more open so they could interoperate together, and so that APIs became more open. It makes a lot more sense, because ultimately, the buy side may not want to have a full tech stack from one specific vendor and instead can mix and match components. On my own side, I suspect that open source offering will become more prevalent when it comes to trading technologies, where offering support will become a key way to monetise these offering. Indeed, that’s one of the reasons I open sourced tcapy, which to my knowledge is the only open source TCA Python library out there currently.


Liquidity and execution

The buy side has many decisions to make when it comes to liquidity. One key question is how many liquidity providers to have. On the surface of it, you might think more is better! However, the panel noted there was a limit. If you have many liquidity providers, in practice, it is likely that many will simply be recycling liquidity. Hence, having a large number of LPs can just end up increasing signalling risk. Whilst markets are becoming electronic, relationships are still important. Of course, there isn’t always going to be a full two way information share between buy side and sell side, given that some information will likely be kept proprietary. There’s also the question of whether to trade bilaterally or on a more anonymous basis via an ECN. One theme which came up again and again, is that relationships really matter not so much when times are good and liquidity is fine, but during crunch times. It was also noted how data can also inform how “good” a trading relationship is.



A lot has changed in recent years during the pandemic, which has affected many parts of society. The FX market obviously hasn’t been immune to that, as evidenced by increased hybrid working. The increased electronification of FX markets means that data is even more plentiful and can be used to quantify market relationships through the use of techniques like TCA. However, at the same time, real life relationships with LPs are still key, especially when times when liquidity is hard to come by.