FX isn’t just any old asset class, you know. It’s traded by corporates, commodity firms, governments, tourists, and private individuals out of their back bedrooms. It’s traded on smartphones, tablets, laptops, in trains, at restaurant tables, during lunch hours and in the evenings. You can trade (or at least buy and sell) the major pairs, and an expanding range of exotics, at kiosks in supermarkets. Philip Beasley-Harling, partner and CTO at currency-specialist hedge fund Blacktree Investment Partners, says: “The FX market is a lot of private contracts. Whoever wants to trade can do so.” So what does this tell us?
First, that not all FX-market participants are qualified, authorised, compliant, thin-fingered financial professionals. Secondly, that significant FX-trading volumes arise because (for example) corporate treasuries have to manage their cash across borders; governments have to (attempt to) stimulate recoveries; passenger aircraft have to refuel. It’s their market too. FX is traded by just about everybody, on every conceivable scale. Multinationals with global wage bills trade FX alongside macro hedge funds with, say, strategies for non-farm payrolls. There are so many different objectives giving rise to such wide varieties of behaviour – all of them competing on a screen near you.
Steve Aldridge, head of global currency and emerging market electronic sales in EMEA at Credit Suisse, says: “There are many factors that influence the FX market; movements may take place for reasons totally unconnected with our clients’ trading activity or the performance of the algorithmic tools we give them.”
Those movements are what make FX interesting, of course, while the scale and nature of those outside factors are much of what make it different. The elephant in the room – there’s always one, although this one looks a lot more like a slowly surfacing iceberg – is algorithmic trading and execution: where FX algos came from of course; how we’re using them more importantly; how they’re evolving most important.
Gary Stone
“We’re finding that clients who trade FX as an asset class behave very differently from clients who trade it to hedge a primary activity.”
Trader and fund manager (and one-time game developer) Tony Manso says: “I use algos to do the majority of my trading for me. They do very well if left alone. However, since they are my creation, I can easily outperform them because I can see things in the market that they cannot.” If this comes close to suggesting that FX algos can serve the purpose of being stupider than we are – okay, but they can be pretty talented pieces of kit as well. Manso puts forward a light heartedly romantic view of FX trading. “I liken it to the olden days where a few brave knights (and a bunch of stupid ones) would go and try to slay dragons. Only a few would survive to tell their story. And of those, only a few would be in good enough shape to go slay another one.”
Our subject for today, in Manso’s terms, is how best to slay today’s and tomorrow’s dragons in order to live happily ever after in the FX markets. First step is to decide on the most effective method of, ah, executing those dragons. Assume (we’ll deal with this assumption straight away) that we’re going to choose an algorithm rather than, say, pick up the phone or join the queue at the supermarket window.
Factors influencing which algorithm to use
Such factors might include, for example, active trading, passive flows, positioning/hedging, et cetera, and yes, it is reasonable to assume that a significant fraction of the FX market is being driven towards the choice of execution algorithms. Paul Tivnann, Head of Foreign Exchange and Commodity Trading at Bloomberg, says: “FX algorithmic order usage over FXGO has increased 30% over the last 12 months as price takers seek to minimize market impact and avoid crossing spreads.”
The elephant in the room – there’s always one – is algorithmic trading and execution
Why the migration? As Tivnann suggests, to execute via algos is to execute quietly and effectively. Joey Horowitz, CTO at Tradepoint Systems, says: “Today the flip-side of not using any algorithm is market impact; there are some algos that are well balanced between making sure information leakage is kept to a minimum while completing the trade in a timely fashion.”
Quiet, quick, effective. And, increasingly, everywhere? Horowitz adds: “It’s becoming ubiquitous. People don’t even realize that they’re using algorithms. They’re using the interface given to them by their firm, and under the cover, when they click on best bid or best offer, they’re really invoking an algorithm.” That might be an algorithm local to the firm, or it might be an algorithm using more advanced technology (co-location, for example) to gain an advantage. “You don’t take an active decision to use an algorithm or not. You might pick a different algorithm, but it’s very hard not to use an algorithm,” says Horowitz. [This throws an interesting light on two other interviews for this feature. One long-term retail FX trader (with a system to promote), speaking informally, said: “Words like ‘algorithm’ make me shiver.” An academic with a specialism in finance, invited to discuss the conjunction of algos with FX, came back with: “They are just instruments, like thousands of others.”]
To borrow a line from a film with “Invasion” in the title: they’re here already. Okay. But what factors drive the conscious choices made by conscious users of execution algorithms? Asked the question heading this section, Philip Beasley-Harling says: “There are essentially two styles of execution algorithm. There are snipers, which are aggressive. They will wait to see what the market’s doing, and if a price comes along that you’ve instructed it to get, it will place an order.” It will, in effect, take the liquidity out of the market without giving away its position. As Beasley-Harling says, the potential risk of a sniper is that a slow connection/fast market might cause you to miss your price. Beasley-Harling continues: “This is where latency is a factor. On the other side, where you’re making a market, you have icebergs. You place an order on the market, and then refill it.” The tip of your iceberg is the first $1 million, say, and when that’s gone, the rest of your $100 million gradually surfaces, million by million.
* The Reserve-Scale Back strategy is a passive algorithm designed to buy on weakness and sell on strength. Example: Buy on weakness 20 million EURUSD with an entry limit of 1.2528. Scaling back 5 pips every time the market takes a slice of 2 million. When price moves away from the algorithm’s current limit, continue to participate by trailing the current market no worse than by 3 pips every 3 seconds (stop before breaching the Top of 1.2540).
** Iceberg (or No-Show) orders only display a portion of the full order amount, including full no-show orders that can be 100% hidden from the market.
Stone says: “Algorithms are a big and growing example of how execution is shifting. Traders are starting to use algo’s as tactical tools. Different execution techniques are being combined. Let’s look at a large order as an example. I could take my order and put it into a TWAP, which will reduce market impact by spreading the execution out. At a certain rate, however, I may decide to buy size because I’m looking for an average result. The algorithm is decreasing risk by working the order so it appears as a series of uncorrelated orders. When the market is at a certain rate, and shows it can handle more size, then I can leverage my bank relationships.”
Note that: different tools are being combined. The main factors that drive a firm’s execution requirements boil down to one (two) requirement(s): getting the order away with a minimum of fuss, on a connection that’s up to the job. We’re not talking about clients deciding at the outset whether they’re passive or active, risk-averse or not; we’re talking about firms actively engaging in the process of matching strategies, and indeed real-time tactics, to algos. Asif Razaq, global head of FX algo execution at BNP Paribas, has spoken in these pages (Q1 2013, for example) about “adaptive” FX algos; users, too, are adapting to a market environment that rewards a sophisticated approach to algo selection. Numbers aren’t everything in this, but consider also the supply side. “Bloomberg now provides clients with access to over 50 algorithmic strategies from 12 liquidity providers on an agency or risk basis,” says Tivnann.
In short: the conscious users know what they’re doing. End of story? Not quite. Algos are everywhere and we’re getting more sophisticated in our use of algos. But we can’t stop here. There are several other factors impacting on the choice of the most appropriate FX algorithm. The first, given that algos are designed, deployed and discussed by people rather than machines, is the human factor.
Rob Maher
“Algorithmic trading is all about transparency.”
How are buy-side clients selecting algorithms?
If our first question focused on the externals, here’s where we look at the thought processes involved in algo selection. Earlier, Tony Manso described the easy relationship he has with his algos. By contrast, Alex Krishtop, trading systems designer, Edgesense Solutions, finds that some buy-side clients have a less-than-clear appreciation of their algos’ finer points. Krishtop says: “It’s interesting that the choice of FX execution algo is often dictated by a kind of vogue. I hear this from clients sometimes, but you can also see it in discussions on the net. Clients don’t always realise the purpose of a particular algo and its applicability to achieving their goals.”
Pressed on his point about “vogue”, Krishtop says: “For example, some years ago benchmark-based algos were on top, including VWAP, TWAP, target close, and so on. Then, for a time, those were superseded by pairs trading in a spectacular variety of forms. Today, the most-discussed execution algos are liquidity seekers.” Hmm. Krishtop goes on to say: “The choice of an execution algo should ideally be made only upon thorough examination of the trading environment and with a clear goal in mind.” True, and one key characteristic of the present environment is the fragmentation of liquidity; that, and the premium placed on liquidity that is readily accessible. “Unique liquidity is the thing,” says Philip Beasley-Harling, going on to discuss the “strange executions” that can arise where, for example, “the same $10 million” turns up on more than one ECN.
The tip of your iceberg is the first $1 million, say, and when that’s gone, the rest of your $100 million gradually surfaces, million by million.
On that point, technology does facilitate the search for liquidity across multiple venues, although it can complicate matters. Putting the case in favour, Philip Beasley-Harling says: “Technology-wise, FX generally is ten years behind equities.” As many of today’s FX traders will remember, ten years ago, they were on an equity desk watching one, or a few, stocks. Beasley-Harling continues: “The difference is, in FX, there are fewer symbols, but nobody owns them. They’re traded in multiple places. There may be twenty, thirty sources of liquidity, and you’re not going to see the best, unique liquidity unless you have a machine to help you do it.” It would be hard to conclude that today’s FX traders, many of them recently arrived from equities, are only chasing liquidity because it’s “in vogue”, but Krishtop does have a point. Krishtop adds: “The decision-making process sometimes requires knowledge and competences that are beyond the scope of the decision maker.”
And yet decisions still get made. Krishtop develops his argument in the box “The Human Factor”. Short version: when the information/expertise required for a good decision is not available, a decision will still be made, based on what Krishtop calls “vogue”, which boils down to what the peer group is doing. Interesting point, and no less valid if we subtract the unspoken assumption that there is necessarily a gap between what’s in vogue, and what really matters. Whether a decision-maker is taking a lead by focusing on liquidity, or following a fashionable concern for liquidity, that decision-maker does seem to be doing pretty much the right thing in current market conditions. Gary Stone says: “The choice of algo depends on the currency pair, the size of the order, the time of execution and market conditions. And liquidity.”
Before we move on, let’s briefly revisit the decision-makers themselves. Gary Stone spoke earlier about increasingly sophisticated FX-algo usage. Similarly, Sylvain Thieullent, APAC Director, Sales, Marketing and Client Services at Horizon Software, says: “We are looking at people who are focusing on sophisticated business; their strategies are not simple. Whether it’s high frequency or not, there is complexity involved.” Comment: such complexity, of course, delivers an interesting audit trail. Indeed, Paul Tivnann at Bloomberg says: “Consistently growing usage of algorithmic orders indicates that buy-side participants are becoming more sophisticated and require more control and transparency of execution. Algorithmic strategies provide access to the pools of liquidity traditionally controlled by the sell-side while transaction cost analysis allows traders to benchmark those strategies to prove best execution.”
And that pretty decisively moves us on towards control, transparency, best execution, TCA. But first – no article on FX and technology, however innovative the technology, however sophisticated the people are becoming who use the technology, would be complete without its section on regulation.
So:
Joey Horowitz
“I have to believe that most users utilise algorithms in some form. Even if they are manually trading they are, more often than not, invoking algorithms under the covers as they decide to release orders to be traded.”
What impact might regulatory change have on FX?
Let’s be practical and innovate. We could approach regulation from any one of several angles. We could discuss the regulation of market participants – venues, banks – in their home countries, and we could discuss current trends towards co-operative international regulation. We could talk about regulated institutions more or less explicitly routing business to regulated venues. Of course, and let’s take that into account. Some banks’ algo offerings do promote their links to specified venues, although it’s also fair to mention that some other venues still claim to be “human-friendly”.
We could talk about all of the above, and no doubt we’d come up with original and thought-provoking insights. Eventually, though, we would have to conclude that although you can regulate many things, the FX market ceased to be one of them when (perhaps even before) Bretton Woods collapsed in 1971. We know that governments lose when they bet against the FX market. Hedge funds win big, but they also lose big. Regulators contemplating the regulation of the entire FX market … can expect to lead interesting lives.
“FX spot is still unregulated, and in my view, it’s pretty much going to stay that way,” says Beasley-Harling. Picking up the same point, Joey Horowitz says: “At the moment, given the current regulations under consideration, I don’t see anything like RegNMS, or similar, happening in FX. SEF rules are not applicable to spot and mostly applicable to price discovery; as opposed to order routing which today is a standard feature of many currently available FX aggregators (whether offered by vendors or built in-house) for most FX asset types.”
Tony Manso
“I use algos to do the majority of my trading for me. They do very well if left alone.”
So let’s not talk about any of that. The more practical, useful, innovative approach to regulation, surely, is to look at the FX market rather than the regulators; to look at, if you like, what their (necessary, of course) work is doing to our market. On this, Joey Horowitz has a point to make. “I believe regulation has actually stifled innovation in equities; it can take a year or more for an exchange to be given approval for algorithmic order-flow or other functional or structural changes. Given the unregulated nature of FX there is a greater potential for new innovations to be generated and explored.”
Horowitz cites two examples. Recently, Thomson Reuters Matching changed their price precision for some currency pairs and EBS Markets announced changes to their inbound order processing timing (through new algorithms for order queuing). “Such changes are near-impossible to make in regulated assets without long approval processes and timescales,” says Horowitz. While it is true that regulatory activity can “ring-fence” (a somewhat discredited term, perhaps, but let’s use it) parts of the overall FX market – FX algos from authorized providers bringing business to authorized venues – it is also true that innovation is wild. The “higher value” of the FX market to the overall finance industry is precisely that it’s as big as nature. However extensive and multilayered the regulatory edifice, there’ll still be space outside for (sorry) green shoots. The innovators will find a way.
Horowitz expresses this rather more neatly. “Maybe the reverse will turn out to be true. Innovation within FX will stimulate regulatory change in equities or other exchange-traded assets. In the end, I believe the FX market is going to drive innovation because it isn’t fettered by regulation.” Which is an original and thought-provoking insight, not least because FX-plus-algos is likely to be even more conducive to innovation than FX on its own. But we’ve strayed onto the wilder shores of the FX market here, and perhaps the time has come to turn our thoughts to control, oversight, reporting, best execution, TCA … the whole “Did we do the right thing?” question.
So:
Our subject for today is how best to slay today’s and tomorrow’s dragons in order to live happily ever after in the FX markets.
How are users qualitatively and quantitatively accessing the added value they achieve by deploying algorithms?
Short answer: in a range of ways. Joey Horowitz says: “Some systems have very good ways to judge results. They include some form of TCA to provide the trader with information by which he can judge which algos are working for him and which aren’t.” Given everything said so far about sophisticated usage, complexity and the overall regulatory background, it’s a fair bet that effective TCA is a selling point. [Horowitz adds one caveat: “Firms have had some success in moving equity algorithms to FX. In terms of TCA, the benchmarking is different. Equities is transparent; you’re able to have very clear benchmarks, and you can refer to them post-trade. In FX, those benchmarks aren’t necessarily there. Sometimes they are, and you can get glimpses of them, but the market data’s largely credit-screened, and it’s unique per client.”]
Getting into specifics, Rob Maher, Global Head of Fixed Income e-Sales at Credit Suisse, says: “When they transact in the marketplace, how do clients evaluate whether they get a good price, whether they have transacted effectively or not? Take a macro hedge fund trading FX aggressively, for example. The benchmark that they will typically use is implementation shortfall, or slippage versus arrival. Ultimately to be successful, they need to effectively manage the trade-off between rapid execution and having significant market impact.”
Analysis is a real-time, at-the-time process as well as part of the overall approach to oversight and control. Maher also cites cases where time isn’t the key factor, but volume; where TWAP and VWAP benchmarks might be applied. Maher says: “After clients have been trading for a while, and we have accumulated a statistically-significant data set, we can go in and make recommendations as to how they refine the choices that they make.” Maher goes on to make the larger point that algorithmic trading, is, as he puts it, “all about transparency,” which implies an on-going virtuous cycle featuring execution, analysis and indeed regulatory oversight. Paul Tivnann says: “Algorithmic strategies provide access to the pools of liquidity traditionally controlled by the sell-side while transaction cost analysis allows traders to benchmark those strategies to prove best execution.”
Alex Krishtop
“Clients don’t always realise the purpose of a particular algo and its applicability to achieving their goals.”
So what does the buy-side look for in an algo?
Final question. Philip Beasley-Harding says: “There are a few characteristics you need for a good algo. One, you need a good source of liquidity. That might be multiple sources – banks, hedge funds, retail liquidity – and they might be of different quality. You need unique liquidity, appropriate size of order on the platform, good-quality credit; you need to look at the contractual obligations on the parties and the speed of the platform. If you know there’s liquidity out there, you need to be able to go and get it.”
Effectively, what makes a good algo is what makes a good execution. This can be about the market as much as the algo. Steve Aldridge says: “Our clients regularly use AES for trading emerging-market currency pairs; spreads tend to be wider and, therefore, the opportunity for spread capture is greater.” Beasley-Harling continues: “Good liquidity, a strong execution platform, and all the back end, the TCA. Finally, there’s the quality of the execution algo itself. How well can you work an order?”
The quality of the execution algo is … you?
The Human Factor
Alex Krishtop, trading systems designer, Edgesense Solutions, argues for rational decision-making in algo selection.
Finding the most appropriate FX execution algorithm is surprisingly often interpreted as finding the best all-purpose algorithm for every situation. The search becomes almost analogous to the search for the Holy Grail, not least in the sense that the object of the search ultimately doesn’t exist. With that in mind, what is the most effective way to select an FX execution algorithm?
Trading systems designers, and indeed traders, should first and foremost be realistic. Your choice should only be made after a thorough examination of the trading environment. The various venues differ in what they offer to liquidity providers and liquidity consumers. Therefore it’s essential to know your counterparties and what is available to them.
Don’t even think about making a choice without a clear goal in mind. As so often, it is possible to improve one metric at the cost of another. This is why we need to fully understand the intention behind every action taken within the framework of the general strategy. With an execution algo, the key question will most likely be the time/price priority: we can fill the whole volume (well, at least in most cases) almost immediately but we will get poor average price – or we may choose to implement an algo that controls the price but is unable to fill the whole volume in a timely manner.
There are many more criteria that ideally should be taken into account when choosing the most appropriate execution algo. Behind all of them is this core consideration: the choice should be made on rational grounds. Seriously. Talking to consultancy clients, I have come to understand that many decisions are made on non-rational grounds, with potentially damaging consequences. Typically, such decisions occur when the decision-making process requires knowledge and competences that are beyond the scope of the decision maker – but the decision cannot be deferred.
If a decision-maker doesn’t understand something, or is unable to do a proper analysis, then he will tend to follow any known model of behaviour that seems attractive to him. It will seem attractive because it is followed by other people, traders perhaps, whose status brings them to his attention. Such a decision-maker is in effect following what is in vogue in the market at any given moment. I have come to understand that a lot of decisions are made on the basis of what is in vogue.
As I said, this tendency brings consultancy clients to me, but It can be observed even from reading discussions over the net. For example, some years ago benchmark-based algos were on top, including VWAP, TWAP, target close, etc., then for some period they were superseded by pairs trading in a spectacular variety of forms. Today, the most discussed are liquidity seeker algos (Crossfire, Iceberg, MOC, etc.). New clients don’t always realise the purpose of a particular algo and its applicability to achieving their goals.
The choice of the most appropriate execution algo might be no less important a task than choosing the general strategy and the trading platform – and requires a rational mind and probably professional advice. Emotional decisions, be they related to entries, exits, risk management or infrastructure have equally disruptive impact on the overall performance.
