It is an old saw in the FX industry that firms respond to customer demand and with algo usage on the increase for firms executing FX trades, inevitably eyes have turned to XTX Markets with the question, ‘will it also enter the market for algos?’ As one of the biggest FX market makers such questions are inevitable and the answer, thanks to the live debut last week of the XTX Execution Algo, is an emphatic ‘yes’.
The algo strategy has been built in house and leverages XTX Markets’ expertise in market making, specifically, as Matt Clarke, head of distribution & liquidity management (EMEA) at the firm explains, “Market making is why we exist as a firm and we are using all our knowledge of market microstructure and trading in this product.”
The XTX Execution Algo is a principal algo which has a very simple outlook, reducing slippage from arrival – or implementation shortfall – but is built upon tremendously complex processes, effectively, as Jeremy Smart, global head of distribution at XTX observes, “We expect the algo to trade much like our own proprietary market making business.”
The new algo is available via multiple channels, including an XTX GUI as well as via FIX connections and the major third-party platforms that support algo strategies. Clients are able to face XTX directly, alternatively the firm is partnering with several selected banking partners to make the algo available to their own clients if preferred. “We think this is a new market structure,” says Smart. “Our partner banks can put their high-quality credit together with our execution expertise and the end client doesn’t have to change anything in their workflow or whom they execute the trade with. They know their orders are being routed to XTX, that is fully disclosed and we want that level of transparency for the client, especially around who is actually executing their orders.”
“We have done a lot of research work over the years aimed at creating the XTX fair value mid-rate. This is our alpha and the new algo leverages that work.”
The partnership approach is likely to help XTX reach clients that were previously perhaps beyond it, not least those corporates and asset managers that are clients of such banks and who make up the majority of market participants using third-party algos. Conversely it also allows those clients to access the expertise of a firm that has established itself as one of the small group of major players in the spot FX market especially.
Maintaining the simple approach to the issue of what is essentially a simple subject, users are able to select one of three settings, low, medium and high urgency, they are also able to establish price barriers at which the algo will either stop executing or switch to high urgency to complete the order. Clients can also change the urgency setting during execution and if using the XTX GUI or a third-party platform that provides view only analytics, they are able to follow the execution in flight. “Clients like the level of information they receive during the execution,” says Smart. “It allows them to see the bigger picture, while the algo deals with the microstructure issues around the child orders.”
Being an implementation shortfall algo it has a very specific objective, namely reducing market impact, and here XTX’s high profile should really help clients avoid excessive impact purely because the firm is so active on both sides of the price. This means that discerning client algo orders from the firm’s proprietary business will be nigh on impossible to other market participants. The aforementioned study of the microstructure of the FX market will also help. “We have done a lot of research work over the years aimed at creating the XTX fair value mid-rate,” explains Smart. “This is our alpha and the new algo leverages that work.
“The algo has a significant amount of information driving it by definition, not least the predictive alpha our market presence and experience can provide,” he continues. “We have done a lot of work on watermarking our liquidity over the years and that has taught us a lot about different channels and different streams and allowed us to protect the alpha in our price. The new algo uses the same IP, thus it offers clients the same protections.”
Segregation and Netting
Inevitably when a major trading firm releases such a product, questions are asked about how it interacts with the proprietary trading business. As a signatory to the FX Global Code, XTX takes these questions seriously and Smart is keen to describe how the new algo is segregated from the main FX business. “We have built our own standalone EMS (execution management system) which is segregated from our market making system,” he explains. “The two do not talk to each other, they merely send skews that are then aggregated for sending to market. By aggregating skews, no information of relevance or size relating to the client order is revealed. The level of urgency dictates the skew we may show, not the size of the order.”
If the market making business and client algo are trading in the same direction the intention is for the former to trade to maintain a broadly neutral position in the market and excess fills from the skew are allocated back to the client algo. “This effectively means that the skew dictates the order fills for the algo,” explains Clarke. “It is the same methodology for two client orders that are in the same direction, we do not look at this on a first in-first out basis, we think that’s fundamentally unfair – consider a 1m order entered a second behind a 500m order – and heightens the risk of information leakage.”
Inevitably when looking at a firm with the market presence of XTX thoughts turn to internalisation, however the lack of an industry-wide definition of what that term actually means is unhelpful. Reiterating that the urgency will dictate the skew XTX shows and not the size of the order, Smart says that the firm is looking at providing a netting solution, but is concerned about the potential information leakage such a solution can bring. “We are looking at different solutions for this, for example minimum amounts for netting,” he says. “We want to make it as hard as possible for firms to fish for larger orders. At the moment, however, we are stressing to clients the benefits we can offer on an individual order basis through our scale and expertise.”
A Different Analytical Look
A very interesting aspect of XTX’s new algo is how, at a basic level, it may be counter-intuitive in how it executes. For those market participants used to using basic TCA tools that look at top of book and little else, understanding the benefits of the XTX Execution Algo will need a deeper, more complex look at transaction cost analysis. Clarke is keen to stress that XTX firmly supports the use of third-party firms to provide execution quality analysis, however he also observes that the analysis needs to again be sophisticated in its approach.
“We don’t believe that just because a fill is passive it represents best execution, buying passively in a market that is about to drop is very easy, but is it best execution?”
“Our alpha allows us to create a different mid than may be obvious in the market,” he explains. “And where our mid is predictive that means that sometimes an aggressive fill can be better than a passive fill. Imagine buying on the bid in a 10/12 market whose next price is 4/6 and then 7/9. With a predictive mid you would want to pull the 10 bid and aggress the 6 offer. We want clients to look at the child orders on a T+one second basis, for example, because that would indicate if the fill genuinely captured spread.
“We think good analysis – and several independent providers can offer this – allows the client to look back with the benefit of hindsight and ask themselves whether it was a good fill or not,” he continues. “We don’t believe that just because a fill is passive it represents best execution, buying passively in a market that is about to drop is very easy, but is it best execution?”
The different approach to analysing what is a growing number of algos on the market also drives XTX’s hopes that the new algo will add to what it says is good initial traction in the market. With so many algos on the market it can be difficult for clients to accurately work out which perform well, and which don’t. Smart observes that third party TCA providers now offer peer analysis, meaning that clients can access pooled (and anonymised) analytics on different individual algo strategies.
Rather than having to execute hundreds, perhaps thousands, of orders to gain the necessary depth of analysis, by accessing the peer group meta-data, clients can gain insight to help them with their algo selection. “We think taking this approach allows clients to make better decisions using more robust data,” says Smart. “We think that performing this analysis will allow the best to float to the top and the objective performance metrics will see algos like ours attract a greater proportion of business.”
In many ways the approach taken by XTX reflects that taken by so many successful products over the years; back up a simple client interface with very sophisticated analysis and technology – as Clarke wittily observes, “Too many algo interfaces require a PhD to navigate. He adds, “We wanted to create a simple solution that clients can use easily and with the confidence that our scale and expertise is driving the decision at micro level. The client is blending their own alpha in terms of execution timing, with our detailed knowledge of the market structure, in an environment in which their workflow doesn’t change.”