Speakers at the Profit & Loss Forex Network New York event highlighted that how you define, measure and respond to market impact when trading FX depends on a number of different variables.
Opening the discussion ‘Market Impact – Finding Execution Styles that Work’ moderator Paul Aston, the CEO of Tixall Global Advisors, argued that while there are a lot of ways to define what constitutes “market impact”, at the end of the day it represents a cost for firms trading FX.
“The way that I look at it as negative alpha. And so if you can recognise it and measure it using more advanced TCA, particularly pre-trade and in-flight TCA, then you can conserve on negative alpha, which is then positive alpha,” he said.
Doug Cilento, COO of Dynamic Beta, concurred, stating that throughout his career managing execution for systematic hedge funds and asset managers, he always focuses on market impact as the primary measurement of trading cost. Indeed, when it comes to calculating the cost of trading, Cilento said that if your execution style involves breaking trades up into small pieces and disseminating child orders through the market via algorithms or other strategies then you should be more concerned about your overall impact as opposed to bid/offer spread that you might pay on any one transaction.
“It’s a key component of understanding your execution costs, which if you’re a systematic fund manager you need to incorporate into your portfolio generation process to know what to expect to pay,” he added.
Jim Cochrane, head of sales for North America at BestX, added another layer to the discussion by differentiating between market impact at a technical level and information leakage, which could in turn cause market impact.
“Market impact is part of our cost model,” he said. “We measure execution rates outside the bid/offer spread as potential market impact, and then we display that in BestX with the aim of helping our clients minimise that spread. However, on an algo trade where every one of your trades might be at or inside the bid/offer spread, that in our equation will have zero market impact, but at the same time it is possible that you’re moving the market against yourself because of information leakage or because of how the algo was deployed.”
“Drinking the Kool-Aid”
David Mercer, the CEO of LMAX Exchange, however, offered a different perspective by arguing that market impact is not always a negative thing for firms trading FX.
“So I know this is a little contrarian to say, but the market has drunk this KoolAid in the last two years that market impact is always a bad thing. I have a bunch of day trading firms who love market impact, they actually want to create it. And if you think back to the old days of FX, if you wanted to sell $100 million the last thing you did was start selling in $10 million pieces, in fact, you probably bought $50 million. That’s the reality of it, sometimes you want the market to react, sometimes you want that information leakage,” he said.
In terms of TCA, Mercer highlighted five metrics for measuring execution that both liquidity providers (LPs) and trading venues should offer buy side firms as standard: the bid-offer spread, their fill rate, the hold time on their trades (if last look is being used), their price variation and their market impact.
“If you’re not measuring those five things post-trade to inform your next pre-trade decision, then you probably shouldn’t be in charge of execution strategy,” he added.
Mercer, a long-term and vocal critic of last look in FX, also argued that this practice makes measuring market impact more challenging because different LPs might operate different hold times on incoming trades and during the difference between these times orders can be broadcast to the broader market.
Cilento responded that whether or not market impact is a positive or a negative when trading FX really depends on the alpha decay profile of the strategy being implemented.
“If you have a short-term alpha decay profile, you’re getting in a trade and hoping that the market moves so that you can get out of the trade at a profit, then you want market impact – that is your alpha. But if you have a long decay profile, or zero alpha, then you definitely don’t want to see market impact. You want as few people to know what you’re doing, you want markets to stay steady all day. So it really depends on your alpha decay and your trading horizon,” he explained.
Similarly, Brandon Primack, head of execution management for the Americas at 360T, pointed out that firms with different trading styles and requirements will have very different priorities regarding their FX execution. For example, he said that many of the firms trading on the 360T platform are non-alpha generating and therefore are not necessarily making decisions based on how many milliseconds an LP might be holding their trades for. By contrast, Primack claimed that many of these firms might be more concerned with the ability of their LPs to warehouse risk.
“When I look at my book and my statistics, the concentration of flow ends up in such a small number of LPs. The rivers flow to the oceans and there really aren’t that many unique LPs warehousing risk that you want to be there time and time again on the other end of your flow. So we focus a lot of our education efforts with clients around this area,” he commented.
The fact that different client types will have different priorities when executing FX is why Kevin Wolf, the CEO of FastMatch said that he tries to let clients lead the conversation in this regard.
“In terms of the framework and the metrics that were laid out, I don’t disagree with any of that. But the reality is as a platform provider and a service provider, we don’t tell our clients what to care about. Yes, we try and educate them and help move the market forwa
rd, so if someone says to us that they care about hold times we’ll help them optimise hold times for their trading, if they care about market impact, we’ll
help them understand that better. But in practice it’s not always black and white for us because, for example, with someone wanting to know more about their market impact the next question is always: well how do you measure it? Because not everyone measures it in the same way. Response times, fill rates, post-trade price impact, reject rates, part fills, spreads, cost of rejects are all important, but not always equally as important to all of our clients. We get all different types of firms using our platforms and over time you see their views and approaches change, and as they do we have to continue supporting them,” he explained.
Cutting through the noise
The panellists then moved on to discuss the challenges associated with the data analytics required in order to preserve alpha, with Mercer arguing that this largely consists of the simple arithmetic of taking the point of a trade and then marking it out by currency pair, time of day and market volatility. However, Cochrane argued that although some of the mathematics involved in simply working out the costs of a transactions are not that complicated, it does get more complex when the analytics look at hundreds or thousands of trades to pick up trends that are occurring in the market after a firm trades. In addition, he claimed that finding and using the right data set to make such analytics effective can also be tricky.
“I would agree, it’s simple but the data is noisy,” added Aston. “You’ve got a cocktail party problem where there’s lots of different sources of information all making noise at the same time.”
To which Mercer immediately responded: “Which part of the data is noisy? It’s not the underlying price. What’s noisy is if you have a myriad of last look feeds because most of that isn’t real liquidity…..People tell me that they have 50 different price feeds, but the fact is, they don’t. They probably have three firm and three decent non-firm feeds, and that’s it. They could chuck the rest away because it’s all reinvented, that’s where the noise is coming in.”
From a buy side perspective, Cilento said that the big challenge around what the other panelists were referring to as “noise” is figuring out the degree to which market moves that occur after trading were caused by your trading or by external factors.
“If I’m buying $200 million dollar/rand and someone else is buying a yard of dollar/rand, am I impacting the market? Are they impacting the market? You need to have a lot of data for statistical significance – to be able to say my impact is 20 basis points. And once I know that my impact is 20 basis points then I can start to unpack what’s driving my market impact and look at what factors I have control over, and I can start to tweak my strategy in order to reduce that impact. But the challenge is statistical significance, knowing that this was my impact and not some other random event in the market,” he said.
Finding patterns in the data
Primack was quick to agree with this last statement, adding that he sees clients doing repetitive trading patterns against a fixed liquidity pool, and that often these patterns start to rhyme.
“So you can start to break out: what’s me and what’s the liquidity provider? What is the last look hold time? Because not all last look is created equal, I think that we can all agree on that. So even without having a full dataset for the market you can dig into the numbers, especially in a disclosed trading environment with a fixed liquidity pool behind it, and start to pull out what I’m doing versus what the market is doing,” he said.
This conversation is exactly why FastMatch launched its consolidated tape for FX, said Wolf, explaining that it is not designed to replace the data set that firms are using to analyse their impact on the market, but rather should be thought of as a supplemental piece of information that can be used to aid this analysis.
“The tape has a role in the world of TCA and we are seeing TCA providers consuming the tape. At $100bn of ADV, the product is still evolving. It doesn’t replace any of the other data we’re talking about, it’s complementary to what already exists,” he said.
This led Aston to observe how challenging it can be to measure information leakage, which he described as a latent variable in what had been discussed previously. According to Aston, looking at transacted data, events that were time-stamped and the prices and volumes associated with that, is “water under the bridge”, whereas what trading firms need to know is how their presence in the market and their interaction with the market is fundamentally changing the behaviour of others.
“So if there’s a configuration of the order book and I go and touch top book or I sweep some liquidity, what is that implicitly transmitting to the existing orders? Are there going to be removals in terms of cancellations or expiries? Or other changes to that order book? If I touch the order book, have I changed the liquidity that I subsequently have to deal on?” he questioned.
Aston added: “Then there’s the issue of market impact coming from the leakage of , for example. So if I deal with a counterparty that’s going to internalise my trade and none of my information leaks out versus somebody who’s immediately exhausting that back to the market, am I actually trading against my own flow? So the minute I actually go lift an offer, is someone else now offsetting that liquidity and then I’ve actually doubled my footprint implicitly because of, the types of counterparties I’m in interacting with? So that information leakage and the the liquidity that you’re seeing is still going to be there, I think is also a very important thing to consider. And this is difficult to measure.”