The Gateway to Algorithmic and Automated Trading

EBS: Equitable Trading

Published in Automated Trader Magazine Issue 22 Q3 2011

Since its original incarnation in 1993, EBS was for many years a bank-only manual trading platform for FX. Then between 2004 and 2006, it underwent two major changes - the introduction of buyside counterparties and API trading. Automated Trader’s founder, Andy Webb, talks to Brian Andreyko, EVP Global Head of FX Product and Development, about EBS’s automated evolution past, present and future.

Brian Andreyko

Brian Andreyko

The changes made to the EBS market between 2004 and 2006 were pretty radical. How did these come about?

They were actually much more progressive than they might appear externally in that a considerable amount of consultation with the EBS owner banks on matters such as non-bank participation and API trading (which we refer to as Ai) was going on behind the scenes since 2003. There was also an extended trial of Ai for banks only during 2004 and EBS Prime was originally introduced just for smaller and regional banks. The key point was that owner and participating banks should feel entirely comfortable with the changes being proposed before the introduction of non-banks to Ai via EBS Prime in 2005.

What is the current constitution of EBS's participant demographic?

Ai has been a growing percentage of our volumes and now accounts for about 50% of our activity. While a significant proportion of that is attributable to buyside non-bank participants, it has been interesting to observe how banks (who account for some 60% of our total volume) already make extensive use of Ai in various areas.

Some of this activity is what I would refer to as aggregation, where although a trader at a bank gains access to different liquidity pools, the activity is still essentially manually driven. Some bank Ai activity is e-commerce risk management, where single bank portals use EBS to offset a residual risk position. Finally, in common with non-banks, some banks also run proprietary trading models.

That sounds fairly close to the diverse order book that many trading venues aspire to?

Well, we don't have any retail or non-professional participation, but I would say that we have an unusually diverse mix on the platform. That ranges from high-frequency Ai proprietary trading, to e-commerce risk management, to manual trading. The mix is also unusual in that we have what I refer to as a very long 'tail' of banks around the world that don't necessarily execute huge volumes, but in aggregate still represent real and natural interest. For example, we have ten banks in Poland who trade the majors on EBS on a daily basis, which adds further diversity to the liquidity mix.

The intriguing thing is that while Ai activity has grown, we haven't lost any manual key stations; we still had some 2800 active key stations in April. The bank demographic has obviously changed significantly in recent years due to mergers among some European banks and events such as the demise of Lehman. At the same time, new banks from Eastern Europe, Russia and the Middle East have joined, which has considerably extended our footprint as well as maintained stable trading volumes (see Figure 1).

figure 1

Figure 1: EBS global ADV (single count) - Jan '96 thru May '11

Which of these new countries/regions have been the most active?

Russia has become quite significant in that we now have 41 Russian banks live on EBS and are developing the offshore OTC markets alongside MICEX. While we obviously have the Russian ruble on the platform we also see noteworthy trading from Russian banks in the majors, such as EURUSD.

The majority of these new banks on EBS tend to be taking keypads for manual trading, but some are also active Ai traders - especially where they are supporting the local currency. For example, we have three banks in Russia that are actively using Ai, but just for USDRUB trading. The platform also offers trading in EURRUB, the Basket Ruble Hedge Trade (BKTRUB) and the RUB1M (Non-Deliverable Forward).

The overall result of these various activities is that we have seen substantial growth in ruble-related volumes (see Figure 2).

figure 2

Figure 2: EBS ruble trading - Jan '09 thru May '11

How did EBS's recent shift to decimalisation come about?

There was a clear trend towards decimalisation apparent on other platforms that was also linked to an ongoing trend towards aggregation. We monitored this situation closely and listened to feedback from a number of regional and smaller banks who expressed the desire for access to an independent platform with decimal pricing capabilities. In response, we decided to implement decimalisation in a phased approach. This started with a pilot involving Commonwealth pairs that began last September. As a result of that pilot, we made a number of changes that were incorporated into the decimal program for our five major pairs that we rolled out in early March this year.

What did you learn from the decimal pilot that prompted changes to the decimal roll out for the majors?

The changes fell into two categories; those for manual traders and those for Ai. As regards the manual traders, we found that they needed additional flexibility around their screen views. Not every manual trader wanted to view the 'noise' of the decimals, but they still wanted to take advantage of them when it came to execution. For example, if they want three digits they can have them, but if they want just a two digit option and to pad the end they can do that too.

One of the most important considerations was that any enhancements around flexibility should not change the fundamental method of operation. We have a lot of users who have been trading on EBS keypads for years and so it was crucial that they could maintain their existing workflow, while still benefiting from the decimal environment.

For the Ai traders using our EBS Live1 feed, we needed to add different levels for the decimal view, as well as introducing a spread and amount view. This allows them to quickly answer questions such as how far down into the order book they have to go to get a certain amount done. So I suppose you could say that in the case of both manual and Ai trading we were trying to improve the 'view'.

What effect has decimalisation had on volume and Ai participant activity?

As regards volumes, indications are positive. Overall volumes are up and the quality of liquidity remains high. Spread compression and depth are the driving factors behind increased volumes. As regards Ai activity, it is evident that some participants successfully adjusted their trading models in advance and that these are already working very well in the decimalised environment. Some other participants don't appear to have made adjustments in advance and we have noticed that their volumes have dropped off, either because the model is not hitting as many trades or because it has been taken off-line for (re)optimisation.

How are you addressing some of the issues, such as quote interrupts, that can arise from API trading?

This has been an evolutionary process for us, starting with our introduction of an automatically enforced minimum quote lifespan (MQL) in 2009. The key point is that this was (and is) applied automatically from within the matching engines; it is not advisory, but mandatory on EBS.

Since the introduction of MQL we've continuously monitored the market and made further adjustments. For example, post-decimalisation we've become rather stricter about top of book quoting behaviour. If people are improving the price by a tenth, the whole idea is that their orders should come into the market and be at risk and tradable; they shouldn't immediately disappear again.

As a result, we now have more demanding requirements around fill ratios and quote holding periods, which we believe are appropriate in order to accommodate the greater number of quotes seen in a decimalised environment. All quotes in major currencies are now automatically held in the market by the EBS system for 250 milliseconds.

As regards fill ratio requirements, if a participant fails to meet the requirement (especially as regards top of book quoting when they are improving the price) we will provide detailed information and analysis illustrating the problem from our market records. This gives the participant concerned the opportunity to fix the problem by a specified deadline. If they fail to achieve this, then we reserve the right to enforce a significantly increased hold in timer (MQL) on their quotes, though this is a final resort - we don't just summarily impose it.

Our insistence on the minimum quote lifespan isn't just arbitrary. Because we are a bilateral market and not an exchange (which we regard as an important distinction) we feel that it is essential for the benefit of the market as a whole.

It's been intriguing to note that our stance on matters such as minimum quote lifespan have attracted industry interest to the extent of it being adopted as part of market practice rules more generally. For example, the Bank of England's London Foreign Exchange Joint Standing Committee (FXJSC) has recently been revising the Non-Investment Products Code (NIPS) which relates to the conduct of wholesale markets including FX. A lot of what is now being incorporated into that is based upon the market practices that we currently apply, such as the requirement for all prices to be at risk and held programmatically (not just on a best efforts basis) in the market for a minimum period.

What's your stance on the question of the 'last look'?

It is common knowledge that certain venues offer certain counterparties the opportunity to go into a match on trade and then have the option to say 'I don't want to do that trade after all because I don't like that counterparty name or price'. As a result, some market participants assume that we also offer that, which we absolutely do not.

This sort of practice (and other practices such as paying to get further up the order book) may happen on other platforms, but on EBS everything is matched in strict price time priority order. From our perspective, it is absolutely essential to provide a level playing field for the equal benefit of all participants, not just a select few.

Despite the growth of automated trading on EBS, you still seem to be investing significant effort in new functionality for manual traders. Why?

As I mentioned earlier, we have an unusually diverse and balanced order book in terms of participants, which we consider essential for a healthy market. Therefore we have introduced manual functionality such as Continuous Match and Pip Discretion to ensure that our manual trading participants are not at a disadvantage in comparison with their automated peers. In the long run, if manual traders (who largely represent natural interest) are disadvantaged to the point that they withdraw from the market, everyone effectively loses out in terms of the quality of available liquidity.

Another consideration is that much of the functionality we have recently introduced for manual traders uses the sort of 'if then' logic typically associated with automated trading models. So in a sense one could argue that it is bringing the manual trader closer to trading in an automated manner.

What is your strategy as regards co-location? How have market participants responded to this?

We offer co-location facilities in the three data centres that house our three matching engines - New York and London, and proximity hosting in Tokyo. I would say that something that has changed recently as regards co-location is the attitude of some banks. Our non-bank participants recognised the value of co-location from the outset, but some banks did not initially appreciate what it could offer. However, the banks have come to realise the benefits associated with co-location, so we have seen an appreciable influx of banks to our co-location facilities in the last year or so.

Does the fact that EBS operates three matching engines and that participants may be running models in co-location facilities remote from a particular engine raise any issues?

We run three matching engines, because that gives the advantage of faster execution at the source of liquidity in regional markets as well as greater system resilience. While this adds some complexity to the EBS architecture and development timelines we feel the benefits outweigh these costs.

As in other areas, our principal objective when managing and monitoring our architecture is to ensure a level playing field at all times for all participants. For example, we have recently stipulated that anyone trading via API in Asian markets must trade within that region during the Asia trading day only. This is to avoid any possibility of the time taken for quote cancel messages to travel from NY to Tokyo causing 'artefact' quotes to remain on manual traders' screens in Tokyo that are not actually tradable. The key is that we don't want to disadvantage the manual trader but rather maintain a level playing field.

Do you offer just the one type of co-location?

Yes we only provide a single level of co-location or in Tokyo, proximity hosting.

Finally, while EBS is commonly associated with spot activity, I'd like to ask you about any other FX instruments where you see particular growth in interest?

I would say that non-deliverable forwards (NDFs) are a good example of that. These are almost exclusively used in relation to a major currency, typically US dollars, versus an emerging market currency that is subject to capital controls and therefore non-deliverable. As a result, corporations use them extensively as hedging instruments for their emerging market activities, but they also attract a lot of speculative activity as well, which obviously assists with liquidity. That is certainly reflected in the volume growth we have seen in NDFs listed on the EBS platform (see Figure 3) since 2009. Although NDFs on EBS are predominantly traded manually, we are in active discussions with some banks regarding Ai price support. We expect that regulatory changes in the derivatives markets will encourage take-up of electronic trading of non-deliverable FX (forwards and swaps), and we are well positioned to support that transition.

figure 3

Figure 3: EBS NDF volumes - Jan '09 thru May '11