Staying Ahead Of The FX Technology Curve

Share

With Christopher Matsko, Head of Foreign Exchange Trading Services, Portware
I joined Portware in August last year and before long, I started digging into the platform and its operation. It became clear that Portware had sound technological know-how: it had all of the advanced execution capabilities for equities and futures, and had applied those execution tools to its FX trading platform, creating a premier multi-asset EMS. Without getting into material specifics, the result was growth in FX: both in clients and in interest in the solution. What Portware needed more of to bring this momentum to fruition was a certain facility with the underlying FX business landscape, to better anticipate client needs and distil various streams of feedback into viable products.
The FX module within our EMS was originally fit for purpose for macro hedge funds and firms with prime brokers. It was also possible to key into Portware’s API and carry out automated algo trading and RFQ streaming programmatically, from client-proprietary black box models. By executing in this fashion with a Prime handling the settlement details, a firm has the freedom not to have to worry about the operational element of the trade.
However, the decision was soon made that the platform needed to expand its client reach and become ‘stickier’ as a true multi-asset enterprise platform, so they took a strategic position to focus on their core client base, namely Tier 1 asset managers. We then talked to these clients to see how we could help them solve their execution problems, and we found that the limitations in executions were purely a result of an operational gap in the real money FX community: front office traders didn’t focus on the complex workflows in the middle and back office. What was needed was a facilitation layer underneath the execution layer.
That multi-allocation space is very complex and takes considerable effort to get right. Splitting a 30-allocation order into five tranches with various execution venues and liquidity pools is difficult, and there is opportunity in the market for platforms that can help simplify and streamline this process.
The biggest challenge we face is the complex nature of the market place: marrying the OMS and its rules, allocations vis-à-vis the banks’ side of the trade, and blocking and splitting out those trades into the right allocations. Our platform is FIX-based but we have to do all the development work to tie the layers together between the banks and the buy-sides. Much of this work is custom to a given client or bank, but this gives us flexibility too.
From a technological perspective we are ahead of the curve in equities with artificial intelligence running automated executions, but this level of automation (interweaving AI execution capabilities) hasn’t yet been applied to real money FX, as many firms don’t envisage AI-assisted FX execution on a big asset manager’s desk. In my opinion, these institutions don’t yet have the same level of trust in machine execution given the inherent relationship-driven nature of the FX market and that’s perfectly natural. Some firms are very progressive and tech-driven and ahead of the curve, but some aren’t and we need to balance that.
The problem of allocations sits behind much of our technological development; taking a large order and slicing it into an algo trade, an RFQ trade, and a bank stream trade and then tying them all back their original allocations is a considerable challenge. We are working toward a solution to that problem. It will be very interesting to see how things progress once these more advanced execution methods can be taken up by the real money buy-side community.
Another area of development concerns fixing trades. Fixing trades have evolved from a very manual process to one that can be smartly automated and given a more mechanical process. Automating fixing trades allows traders more freedom to focus on and manage other large risk positions in their book. In other words, automation helps eliminate noise or distractions for trades which might otherwise be considered a “nuisance.”
Even though there have been serious issues around the WMR, it is still being used to get large block positions executed at a defined mid-rate (plus commission). Unless a firm is focused on generating alpha or any type of return on those executions, then the best way forward is to simply automate those trades, as quite often asset managers don’t realize the fixing trades are already going into algos as is. All Portware needs to know are the times and the bank pools and rotations and fixing trades can all be done automatically.
Trends in the market
Bank algos present an interesting issue with a diversity of opinions among the larger buy-side firms. Some buy-sides advocate the usage of bank algos as they feel they get a stable rate across diverse liquidity and can hit benchmarks, whereas others say they don’t want to give the bank too much information as the banks ‘can do whatever they want with it’… so we have to try to provide access to both solutions and the challenge is in finding the balance.
Given the relative unease or mistrust in the way trade information is handled by market participants who control the most FX volume, Portware began to consider possible solutions for that problem. We proposed diversified liquidity with minimal information leakage. As an example, we allow clients to formulate their own unique electronic communications network (ECN), combining relationship-based RFQ and streaming liquidity as well as access to third-party anonymous ECN’s (think FastMatch / Hotspot), as well as non-bank LPs. A large asset manager can then cut a large position into smaller slices and work them into various liquidity channels at their discretion; this helps control information leakage and improves executions, which is as much about taking control of that information as it is about transparency and impact on the market. The best way to control market impact and information leakage is to control how much information a firm divulges to a liquidity pool. The ability todiscreetly interact with liquidity channels is the same as the ability to control the flow of information into the market place.
The traditional RFQ model is in slow decline because traders want to have more discretion in trade execution rather than just clicking a button and leaving execution to a third party’s black box. Trades can now be automated according to a given desk’s rules of engagement, and traders can focus on the less liquid, more difficult trades on their blotters. The ongoing trends in FX are a natural evolution of the ‘more with less’ process and automation trends across the wider industry. Firms are being forced to narrow their specialization into core businesses, and they need the rest of the process to be as automated as possible. The issue of FX TCA has been around for a long time. In a fragmented marketplace there is ongoing debate over what TCA actually means, and because there is no “volume-at-price” data in the FX market (as well as other OTC markets), benchmarking and analytics run the risk of becoming more subjective. Despite this risk, it is still the case that the best way of finding out the cost of trading is by including a firm’s own data set in any analytics, in addition to third-party market and reference data. This is as true in FX as it is in equities and any other asset class. Now, there are challenges specific to third-party data in foreign exchange (e.g. validating execution size, credit relationships, anonymous vs. relationship-based executions, etc.), but we should not throw the baby out with the bath water. It is always best to use a healthy mix of market and reference data and proprietary execution and benchmark data, as this gives the firm colour from the marketplace alongside comparisons to their own mid-rate and their own reports, which are best for measuring a firm’s own improvements over time.
So, for example, by combining the results of recently executed trades and some historical data, a firm can view which banks gave optimal spreads at a particular time of day in their traded currencies based on their execution methods; RFQ, algos or streaming. The next logical step is to take those inputs, apply pretrade analytics given current market conditions, and put a smart order router (with an AI wrapper) on top of it to execute a trade at the lowest cost given the characteristics of the order. With a good data feedback loop, a trader can more intelligently feed trades into the market given their size and the market liquidity conditions.
The future of the FX market will incorporate automation, artificial intelligence, and TCA, much like what is under way in equities and other asset classes. It’s about creating a common thread over the entire life cycle of a trade, and of trading holistically, so that the results of one trade inform the next, and a firm can constantly tweak and improve execution. Nobody is quite there yet but change is coming, and this is where the future of cutting edge development lies.
We’d love to hear your feedback on this article. Please click here