FXAlgoNews talks with Alexis Laming - Why a central banker is using FX algos

Alexis Laming is a senior FX trader at the French central bank, Banque de France. He is on the desk which is responsible for FX execution and the implementation of electronic & algorithmic FX strategies for the bank and so we asked him to tell us more about why a central bank is leveraging these toolsets and how they go about it.

FXAlgoNews talks with Alexis Laming - Why a central banker is using FX algos
Alexis Laming

Alexis please tell us a little about your career and background in FX.

I started my career in Geneva as a precious metals trader before moving towards FX in 2009 working for a private bank and then a hedge fund. I joined Banque de France in 2016 to implement an electronic trading solution.  My first contact with algo trading took place in 2009 when a bank showed me a tool to automatically delta hedge my option book. At first I was very skeptical but I changed my mind in a matter of minutes when I understood how powerful these tools were! I never stopped using them since.

We have used algos in our trading for a couple of years now
We have used algos in our trading for a couple of years now

Banque de France is an active member of the Eurosystem, with a strong presence in FX markets for both reserve management and institutional client services. Please tell us a little more about your business.

BdF FX transactions serve 3 purposes: FX interventions on behalf of the Eurosystem and executed by the national central banks; portfolios rebalancing in the context of FX reserves and executing transactions for foreign central banks or international organisations.

What does your job and day-to-day responsibilities usually involve?

We are an execution desk. We receive orders and cover the flows in the market with a double objective: best execution and market impact minimisation. Products traded on the FX desk are mainly spot, but we do have an interest in swaps and options from time to time. On top of that, as a central bank, we are often involved in international study groups with our peers, through the Bank for International Settlements, when it comes to market structure and financial stability. For instance, we have co-chaired the Market Committee study group that has published the report:  Monitoring of Fast Paced Electronic Markets in September 2018.

We work very closely with Banque de France’s Innovation Lab
We work very closely with Banque de France’s Innovation Lab

How did you go about implementing your electronic trading platform and what key features and functionality has it been designed to provide?

The main idea for us was to stay ahead of the curve to be able to better serve our clients with a double target in mind: best execution and market impact minimisation. But they are now multi-dimensional: best execution is no longer calling up 2 banks and trading at the best price! You need to look at 50+ venues, take into account the FX specificities such as last-look and so on. Measuring the market impact is even trickier: mirage liquidity can be misleading. Data and tools are the key to that.

We receive orders and cover the flows in the market with a double objective: best execution and market impact minimisation
We receive orders and cover the flows in the market with a double objective: best execution and market impact minimisation

How long has your trading team been utilising FX algos and what are the important things you are trying to achieve with them?

Algos are very interesting tools, but as all powerful tools, they should be used with great care. You need to understand extremely precisely how algos will behave. Otherwise you can get a result that is the exact opposite of your initial intention. It may seem trivial, but as the market is not fully standardised yet, it can be very tricky. And, you need to be very careful about the parameters you use too.

For us, the main objective when using algos is to diminish market impact and try to manage the information leakage that could sometimes be an issue for institutional investors like us. And algos are great at that, often better than humans. If and only if, they are properly designed and used.

We need to better understand the interactions between algos which may occur through negative feedback loops
We need to better understand the interactions between algos which may occur through negative feedback loops

What types of FX algos (TWAP, VWAP etc) do you currently find most useful and what factors usually determine how you integrate and deploy them as part of your execution strategies?

It really depends; we don’t have a hard-coded rule about the use of algorithms. Contextualisation is the key and this is where humans are still better than computers. As I said, depending on the set of parameters you input, some algos can go in very opposite directions. So, a rule saying “above x million, use a TWAP with parameters a and b if time of the day is between t1 and t2” is very sub-optimal to me, at least with the tools available today.

However, we are in the good position to know how quickly these tools are evolving. For instance, we can see some expected cost-analysis tools being offered by some banks that could be helpful in the execution policy decision. But, I expect that humans will still be needed for this part for a few more years yet.

How do you source the algos you use and what are the differentiators when choosing one provider over another?

We have used algos in our trading for a  couple of years now and, for us, with regards to our very specific position in the ecosystem, we have a preference to use our own algorithms. Mainly because we were facing an issue about understanding in very specific details how the bank algos will behave given a couple of market situations. As a central bank, market stability is really important. So, working very closely with our IT teams and external providers, we have developed our own strategies, which are very defensive ones.

Without data you really are lost nowadays
Without data you really are lost nowadays

How you go about measuring and evaluating the effectiveness of the FX execution algos you deploy?

Data, data and data! I am sure this is not breaking news for your readers, but without data you really are lost nowadays. Every algo execution goes through a very specific evaluation process with a couple of benchmarks that are calculated so we can rank them and assess the effectiveness of the trade.

This is not specific to algos, it is also true for more traditional executions. We want to be able to document the best execution process for every trade done on the desk and update our execution policy afterwards. This is an ongoing process and it helps us keeps pace with the market structure evolution.

Are you using FX algos for every trade or only for specific situations, for example executing a larger block trade?

Definitely the amount to trade is a large component of the decision, but there is no specific and “dogmatic” rule. We adapt the tools to the market context and the goal set for each flow.

In what ways would you like to see current FX trade analytical solutions being improved and how would this help your efforts to achieve best execution and more efficient workflows?

Trade analytics are definitely very helpful. We see that the FX market is moving towards more transparency and that is great. I don’t know any other financial markets that have moved as quickly as the FX space towards more clarity: global code, TCA democratization etc.

We maybe were a bit late and there is still some work to do but the moves we have been seeing are definitely going into the right direction.
Are there any specific things that you would like developers and providers of FX algos to put more focus on? For example by giving you better insight into the venues and liquidity providers that they route to and showing you performance data on these.

Definitely. Transparency is good for everyone. As we have seen with the global code or the increased use of TCAs, the FX market is moving towards more transparency and that is definitely an improvement. I am sure that more innovation will come in the near future on transparency around algorithms.

How closely has your trading team been working with Banque de France’s Innovation Lab and in what ways have they helped to enhance your FX trading operations?

Very very closely. It has really been a team work, in an agile mode. When you want to implement a market solution, you need the market knowledge but you won’t go very far without the IT. And our Lab has this very specific capacity of coordination between the users and the IT while bringing in their specific knowledge. I have to admit that some discussions have been intense! But always with the same common objective: deliver the best possible product given the timeline and budget constraints. And if we have achieved that, it is definitely because of the coordination driven by our Lab.

And it is not over! For instance, we have a lot of common projects going on about exploiting the data and I am sure that we will find a lot of other topics in the coming years.

Do you expect to see more regulatory oversight and control over algorithmic FX trading and why is this important?

Regulation in FX is always tricky because the market has evolved as a non-regulated market. As the French central bank, we are looking closely at financial stability. So definitely, when we see flash events in very developed market currencies such as the JPY, it troubles us: both as a market participant and in terms of financial stability. But the answer is not as clear as it could be in other markets such as equities. For instance, circuit breakers that are effective on exchanges may not be really suitable for FX markets.

We need to better understand the interactions between algos which may occur through negative feedback loops: One algo changes its usual behavior, making other algos believe that something unusual is happening, forcing them to step off the market, and reinforcing the loop by modifying the market. Algos must cope better with this and one way to achieve that could be through more interactions between them.

For instance it could be interesting to observe the behavior of algos launched simultaneously in a simulated market where they could learn from each other. Better understanding the respective reaction functions of algos should promote a better functioning market.

Banque de France is clearly a technology innovator. What are your views on how quickly new technologies like AI and Machine Learning will find new uses in FX applications including algorithmic trading?

Innovation is definitely in Banque de France DNA: when it comes to markets but for our other missions as well. For instance, we run the first interbank blockchain in France, MADRE, for which we have been awarded an innovation prize by revue Banque.

We are going down that route as every significant market participant should. Innovation for big institutions like us takes time; this is why you need to be ahead of the curve to keep the pace.

What do you think will drive increased use of algorithmic FX trading toolsets over the next few years and do you expect to see Banque de France making more use of them?

Transparency is key. More transparency will help algos end users better understand the benefits and risks associated. This should reinforce algo prevalence in every trading desk toolset. And if it helps our goals of reduced market footprint and improved service to our clients, we will definitely keep a very close look at them.