1. Home
  2. The Big Conversation
  3. Episode 1: How are algos shaping the world of finance?
empty image

The Big Conversation - Deep Dive Interviews

Episode 1: How are algos shaping the world of finance?

In the first episode of our new Deep Dive Interview series, we take a look at algos. They have become an increasingly important part of the trading landscape. The proliferation of exchanges often trading the same underlying asset, plus the need to be able to trade across multiple asset classes at the same time has seen demand for these products surge.

Medan Gabbay, chief revenue officer at Quod Financial outlines what algos are and what they do, before Alexis Laming, senior trader of Banque de France, the country’s central bank, explains the user case for algo products, as well as some of their benefits and their pitfalls.

  • Algos have become an increasingly important part of the trading landscape. The proliferation of exchanges often trading the same underlying asset, plus the need to be able to trade across multiple asset classes at the same time, has seen demand for these products surge.

    Detractors worry that this is another arms race that gives the largest institutions an edge over smaller companies and individuals, but improvements in technology have been democratising finance, and the developments in algos should be to the benefit of all market participants. Medan Gabbay, chief revenue officer at Quod Financial outlines what algos are and what they do.

    Algos are different things to different people, so it’s important not to categorise them too heavily. Generally speaking, an algo is the use of some kind of automation for a specific trading objective. It could be to achieve a certain execution style, it could be to minimise impact, it could simply be to automate a workflow. But it’s very important to remember that algos are not the same to different types of institutional clients. A buy-side will have a very different view of what an algo is to a bank.

    I think algos are certainly being used more, and I think it’s important to understand, when thinking about algos, what is their purpose for different types of institutions? So, as the trading environment has become more complex, more volatile, the number of participants has increased, and the rate at which you have to interact with markets has become faster, traders are often relying on automation in general, whether that’s an algo or otherwise, to try and ensure they can achieve a certain objective.

    One of the really important things to remember about the capability of algos is that they’re designed to have an understanding of markets, to be able to use historical data beyond the gut feel of a trader, but to be able to feed that information into making some kind of more informed decision, based on signals and patterns that maybe the trader themselves can’t recognise.

    We’ve certainly seen a shift in different parts of the market where people are relying on algos perhaps to automate their workflow, so if they have an awful lot of trades and they want to take a segment of them and just place them with an algo they know they’re running while they can focus on, perhaps, more complex orders or more complex trades where they can add more value.

    The other aspect of algos also is you get an element of confidence in the outcome. So perhaps in certain market conditions you don’t trust your gut or you want to be able to rely on a certain methodology or pass risk to a third-party broker by using one of their algos. There’s certainly been a big shift as well, but we’ve also seen a lot of limitations.

    The question of are they taking over, there’s a huge challenge around knowing is the algo actually performing as it should be? Do I get the type of execution that I need? And can I even classify what I’m trying to achieve correctly with an algo? Do I want certain market impacts? Can I set the values correctly?

    I think algos are freeing certain groups of people to be more confident in how they’re interacting with different markets, so there’s an element of democratising trading, but at the same time if you’re giving that responsibility to an algo provider and you don’t have a full understanding of how that algo’s interacting with the market, it’s also moving that liquidity to the algo provider and giving them a certain influence over your own trading capabilities.

    I actually think the way the market’s started to move is that institutions are wanting to be able to make use of their own algos, configuring themselves, and directly. So it’s gone from little algo use to third-party algo use, and now people are looking to implement their own technologies that mean that they can precisely manage the behaviour of the algo. So that democratisation of algos and trading is actually being done by the EMS providers, where institutions are pulling algos back into their own architecture.

    I think the main area of growth that banks and non-bank market-makers are really tapping into is this concept of multi-asset trading. It has been an untapped part of the market, partly because most desks are limited in P&L to only one asset class. So if I’m performing a certain activity in equities, it really doesn’t matter to me whether I can add an FX component or add a bond or any kind of derivatives component to that trade.

    What the algo allows is it allows you to systematise behaviour across desks. So it’s the first step in automating across different desks and different P&L, and being able to squeeze that extra profit out of the different trading architectures that you have, without necessarily unifying the systems. Because the challenge that a lot of institutions have is they don’t really have a multi-asset trading architecture in the first place. So unless you’re using external algos, it’s very difficult to look at P&L as a cross-desk concept. It’s certainly the way of the future.

    And again, looking at non-bank institutions like XTX, who very much started in FX, have grown as a massive SI in equities, which now represents a huge percentage of their business in general, that really highlights the benefit an institution can have from using automation across different asset classes. And banks are certainly moving in that way as well. So yes, we absolutely see that.

    The most powerful statement you can make is everyone is using these algos now. It has suddenly become accessible to every part of the industry, whether that’s the buy-side specifying that they want certain algos from certain providers or saying in their instructions that they would like it handled by an algo rather than an individual trader, all the way through to what were previously high-touch desks who are certainly, themselves, moving towards using algos as well.

    So every participant in the market is becoming more and more reliant on automation in general. Going back to the very first question, an algo can take many forms, but we are seeing, particularly in the tier-two institution market, which is the primary market that we serve, so clients such as Kepler, clients such as the Central Bank of France, that algos are increasingly becoming a core of their business. So it really is crossing all segments.

    Alexis Laming, senior trader of Banque de France, the country’s central bank, explains the user case for these products, as well as some of their benefits and their pitfalls.

    The user navigates a very fragmented market. OTC markets are very fragmented these days, and so algos can help the user to hit the best liquidity available on every liquidity venue that they are. Algos are helping users to optimise their execution and to meet a best execution objective that they may have.

    From a market-wide perspective, we have done an intense study with the BIS last year. The work was chaired by the SNB, the Swiss National Bank, by Andréa Maechler, and the findings that this report highlights are two main interests in terms of market structure. The first one is that the price-discovery process looks to be better with algos than without, because mainly algos are slicing orders into smaller pieces so they help the market to absorb big orders.

    And the second point is that algos can help transmit information from one liquidity venue to another. So like in physics you have the quantity and the speed, and the speed is more important than the liquidity. So overall, the net impact of algos on market structure looks to be positive.

    Algos create risks. Well, at least needs, in terms of HR, they need to understand very well how the algos are functioning, which can be tricking in some cases. And algos are good because you have the data about the execution itself, but you need to have the ability to understand and to compute and process the data, to clean the data and to understand and to perform some good analysis.

    The second big issue from a user’s perspective, I think, is it’s really hard to disentangle the performance of the algorithms compared to the usual market noise. How can I tell my boss that I’ve been good at picking up the right algorithm at the right time or I have been lucky? And the question is there, I think. You need to have a very big data set of execution to be able to really understand and disentangle the algorithm’s performance.

    When the market is malfunctioning, because of some external events, for instance, the dynamic algo will probably move out from the markets, and so it will increase the volatility and lower the liquidity available. And so other dynamic algorithms may react the same way and increase this volatility or this market malfunction. So for some of these reasons, sometimes the algorithm’s ecosystem, in some sense, are creating a reinforced feedback loop, and so that can trigger sharp price moves.

    As always with automated tools, humans are not happy to see them malfunctioning. So we understand that a human can make a mistake, fat fingers in the markets have happened since markets exist, but when it comes to an algorithm or an electronic tool, it’s not the same. We don’t want them to malfunction.

    And the second point is that, overall, when it comes to algo execution, algo usage, there is a trilemma about the certainty of execution, the market risk that you want to risk, and the spread you want to pay. And so between these three goals that you may want to optimise when trading and using algos, you can’t optimise the three objectives at the same time. You have to choose two over three. And so, a perfect algorithm, to my sense, doesn’t exist yet. And perhaps the future will prove me wrong, I hope so, but I see it’s really hard to create.

    Are algos replacing the role of traders? And the answer is absolutely not. The use of algos is there to accommodate more complex market conditions, it’s there to accommodate the fact that there’s an increased need for automation because there are higher and higher volumes and smaller profits that have to be squeezed out.

    So using any kind of automation to improve profitability is key, but it’s still the traders who have the knowledge of intent, they have the ability to interact with markets and see wider market conditions in a way that algos really can’t. Algos are good for handling complex problems, but humans are very good for applying thought into how do I need to apply this algo, when is it appropriate?

    The next major revolution in algos is pre-trade awareness of how they will perform in certain conditions, so forecasting, so that when a client calls a trader, he can say, okay, this is what’s going to happen every the next 15-minute segment. And the next thing is this idea of using back-testing to evolve the parameters, so this kind of self-learning component around algos. The real evolution of algos is desks understanding better how to use them and how to deliver benefit for their clients through their use.

    Algos are in a state of constant development. Although technological advancement has been a driving force, the input of human operators remains an essential element in deciding which algos to use, under which circumstances, and for the evaluation of their performance in order to refine them for future use.

    The perfect algo doesn’t exist, and the speed with which they adjust to certain conditions can drain liquidity from the market at key times. But overall, market liquidity in terms of depth and bid-offer spread have generally benefited from these products, and they’ve certainly helped traders and investors navigate an increasingly complex world of multiple exchanges and multiple assets.