The Full FX Temperature Check

In this occasional series, The Full FX talks with leading industry figures to find out what they are focusing on and what they think about the hot issues in the industry. In this episode Stephen Flanagan, global e-FX risk manager and Arnaud Floesser, head of Core e-FX trading at JPMorgan talk algos and last look.

Have there been any lasting changes from the pandemic period in the first half of 2020?

Stephen Flanagan: We have seen another acceleration in the adoption of technology. In the early 1990s there was the introduction of ECNs to FX markets and that was followed by single and multi-dealer platforms around the turn of the century, but it was an osmosis, development was steady rather than quick. The GFC saw that technology adoption accelerate a little but the pandemic really kicked it on.

Many manual voice trading desks around the globe were suddenly dispersed with ‘work-from-home’ orders. However the FX trading market structure showed incredible resiliency and the long term investment in technology really shone through. What was interesting, however, is how quickly some businesses, whom may not have before, embraced technology and with it their algo usage to access liquidity.

Arnaud Floesser: Without doubt the work from home environment encouraged people to try algos, and because they worked well throughout 2020 they became sticky. It is not just about dealing with volatile conditions, though, when markets were calmer the algos performed just as well.

This means, in our view, that the shift may be more permanent, we think a lot of this flow may not go back to the click-to-trade world.

“There is a misconception that additional hold time allows liquidity providers to provide tighter spreads, and the implication is that by not having the delay on last look spreads will be wider, but we disagree with this view. We don’t do hold times.” 

Stephen Flanagan: Without doubt the rise of passive algos has stabilised the liquidity landscape and that means we have a well-defined top-of-book and there are traded points at every part of a move. This means there is generally more predictability to moves, which benefits the broad market.

Another sub-plot with the use of passive algos is that people are now even more aware of mark outs and the effect of their trade, or trading style, on the market.

Has that changed the risk management role?

Steve Flanagan: Temporarily perhaps, because the analytics and tool kits that Arnaud and his team have built allow the trading business to better analyse markets in a shorter time window. Otherwise, not a lot has changed, although we watch mid-points a lot more than we ever did in my voice trading career!

As understanding grows, is there demand for different algo strategies?

Arnaud Floesser: Probably the other big trend in the algo space has been from clients who were already using algos – they started demanding more dynamic and adaptive solutions. This was largely a result of the period last year when markets and liquidity levels were so volatile. Experienced users were cautious of deploying vanilla algos because they could cause increased mark outs and didn’t adapt on-the-fly to changes of liquidity. That led to a new generation of adaptive algos in the industry.

Stephen Flanagan: The algos have to reflect the skill set of a trader and a trader’s main skill is understanding market conditions and adaptiveness. A lot of the development around algos is the result of collaboration between Arnaud’s quant teams and the trading desks. The concepts are put out there, the trading business offers its insight, the products are then developed and they are then tested by the trading desks. Feedback is provided and, if necessary, tweaks and upgrades are implemented before the algos get offered to our clients.

Arnaud Floesser: Understanding liquidity is critical, we conducted a multi-year project to look at how best to unlock and access real liquidity, both for our trading desks and our clients. The resulting piece of technology is now wired in our new algo engine. Market fragmentation is felt by many market participants still.

Liquidity venues are in different data centres, have different network latencies and use different rules. We factor it all and the algo engine then optimizes the way orders get sent to maximize the probability of being filled.

So there are different views on last look depending upon what side of the business you sit?

Arnaud Floesser: Our spot FX market making business has been at zero additional hold time for many years, across all channels and all clients. We do run checks before accepting a trade to ensure the order was submitted correctly, to do a price match within a symmetric tolerance, to validate credit and to confirm some operational and regulatory validations, but other than that there is no arbitrary hold time delay. We do this as fast as we can and don’t delay to see what happens next in the markets. This whole process only takes a few milliseconds typically.

There is a misconception that additional hold time allows liquidity providers to provide tighter spreads, and the implication is that by not having the delay on last look spreads will be wider, but we disagree with this view. We don’t do hold times and we believe our spreads are some of the most competitive for our clients. We strive to offer a fair and realistic interpretation of liquidity in the FX market.

Stephen Flanagan: As we see more algo usage the focus on last look delays may become a bigger topic for clients. Whether these delays are helpful may be increasingly debated.

“Without doubt the rise of passive algos has stabilised the liquidity landscape and that means we have a well-defined top-of-book and there are traded points at every part of a move.”

So more clarity on additional hold times would be welcome?

Arnaud Floesser: I think it would and it would mean people are inclined to show real interest in the market. If you look at all the liquidity we can see as available in the market, for example in EURUSD, it would be close to choice for very large sizes, but the crucial factor is whether you can access that.

As mentioned earlier, we invested heavily and did the math to optimise how we access liquidity, and achieve high hit ratios for our clients. Ideally, realised executions should be comparable to the visible liquidity.

We continue to observe in the industry an ever growing emphasis on quantitative analytics, increased spending on technology via co-location of servers and better hardware. All of this should imply that e-trading can operate without the need for delayed hold time, and that e-trading can deliver high fill ratios.

Stephen Flanagan: I also think internalisation levels lessen the need for this delayed last look – why is it important what the market is doing if you are not going to go to market? It’s also important to point out that the more stable liquidity landscape I mentioned earlier is built upon no delayed last look – by offering clients real liquidity they are accessing a predictable and stable execution environment.

This is also the case down the liquidity curve, I think another lasting impact from last year has been better depth of market.

Has the client relationship changed at all?

Arnaud Floesser: We see trading as a scientific field so we have to take a more scientific approach to it and that does spill over into the client relationship. There are lot more analytics available to drive our conversations these days, and we also see this change on the client side. This often leads to very rich conversations frankly.

Stephen Flanagan: The good news is that many are moving in this direction and that means that when you have discussions with clients you now start to hear them give data back to you, which is great, because it means you both understand the drivers and it helps you go back to the team to present the client’s viewpoint. How we interact with each other has changed for the better.