Graham Capital Management

Research Note, June 2021

Execution in Spot FX
Sanjith Ponnamaneni1 , Chris Jones2 , Ed Tricker3

Abstract
In this paper we highlight some of the mechanics, nuances, and pitfalls unique to OTC FX trading and propose that
building a bespoke FX Aggregator is a way to overcome some of these challenges. We also highlight some of the
considerations that are needed in such an endeavor.
1 Director- Smart Execution
2 Managing Director - Portfolio Solutions & Research
3 CIO - Quantitative Strategies

1. Introduction high access fees and quote restrictions in primary markets. In

addition, there are multiple aggregators (e.g. Bloomberg) which
For most CTA and Systematic Macro investment managers, For-
gather quotes and liquidity from other sources such as banks and
eign Exchange (FX) is a core market, along with equity index
non-bank institutions (e.g., XTX). Banks also have access to the
futures, bonds/rate futures, and commodity futures. However, FX
interbank market, where they can transact significant volumes
is unlike these other asset classes in as much as it is typically
at tight spreads. Figure 1 shows the distribution of FX volumes
traded through OTC forwards rather than exchange-traded futures
across different channels.
due to the relative depth of OTC FX market compared to the less
voluminous FX futures market. This does, however, present a
challenge given the fragmented and nuanced nature of the OTC 4. Liquidity Provider Mechanics
FX markets, and specific expertise and systems are needed to
A liquidity provider (LP) in spot FX is very similar to a market
overcome these challenges.
maker on a futures or equities exchange. They make markets
This paper explores the nature and nuances of FX markets, by providing a two-way quote and profit primarily by capturing
such as fragmentation by various types of liquidity provider and spread. They typically hold the position for some time, looking
types of quote. Furthermore, we show how a carefully constructed for an offsetting trade from other clients or offloading it into the
FX quote aggregation system can overcome many challenges market (if the trade is risk-increasing or loss-making for them).
presented by market fragmentation and enable efficient execution.
A LP quotes a bid and ask price individually to each client,
which is derived as follows:
2. FX Spot Vs Futures
1. The LP establishes a mid-market price using prices from
Futures are exchange-traded and therefore have some advantages primary market, ECN, and interbank quotes. These quotes
such as access to the exchange limit order book and protections are fed into a proprietary model which provides a mid-
from market manipulators. The FX market, on the other hand, is market estimate.
highly fragmented with no central exchange. Price discovery is 2. Half spread is applied to either side of the mid-market
based on the relationships with various market participants and price, and resulting quotes are sent to the client. The spread
the quotes received from each counterpart. This makes execution used depends on many factors (currency pair, volatility,
in FX more challenging, but with the potential for relatively low individual client profitability – this last factor is identified
cost of trading due to tight spreads and high liquidity that helps by a LP using mark out analysis and discussed in more
make it a compelling choice over FX Futures. detail in later sections). Customers providing the most
profits can usually expect to receive the tightest spreads.
3. The FX Market Place 3. Based on the current position of the LP, the mid-market
Trading in FX markets is estimated at $6.6 trillion per day with price is adjusted to encourage or discourage flow on a
Spot FX at about $2.2 trillion/day 1 . However, liquidity in Spot particular side (this is known as skew). For example, if a
FX is highly fragmented. At the top (by volume) are the two LP is long in EUR/USD and wants to reduce their position,
primary markets – EBS (G10 pairs) and Thomson Reuters (fo- they will reduce their offer price (thus encouraging buy
cused on Commonwealth pairs). There are several electronic orders from clients).
marketplaces (ECNs) with HOTSPOT and FASTMATCH among
the more widely used. ECNs have become more popular due to
1 BIS Quarterly Review, International banking and financial market devel-

opments, Dec 2019. Available online at https://www.bis.org/publ/

qtrpdf/r_qt1912.pdf.
 seamlessly adjust and redistribute financial exposures.
The 2019 Triennial Survey shows that the share of FX trading executed Execution in Spot FX — 2/5
electronically edged up only slightly since the previous survey (Graph 3, left-hand

How were FX trades executed in April 2019?

Percentage shares in total turnover Graph 2

Broad voice and electronic execution methods1 Breakdown of electronic execution methods
3 (+1)
50
15

13 (–2) 28 (–5)
40
13
30

27 (0) 13 20
29 (+6)
10
16
0
Electronic direct: Electronic indirect:
2 2 4 6
Voice direct Electronic direct Undistributed Single-bank platforms Disclosed venues
3 3
Voice indirect Electronic indirect Other5 Anonymous venues7
1
Change in percentage points since the 2016 Triennial Survey indicated in brackets. 2 “Direct” refers to trades not intermediated by a third
party. 3 “Indirect” refers to trades intermediated by a third party – either a voice broker or a third-party electronic platform. 4 Single-bank
trading systems (eg Barclays BARX, Citi Velocity, Deutsche Bank Autobahn, UBS Neo). 5 Other direct electronic trading systems (eg direct
electronic price streams). 6 Multi-bank dealing systems that facilitate trading on a disclosed basis or that allow for liquidity partitioning
using customised tags (eg 360T, EBS Direct, Currenex FXTrades, Fastmatch, FXall OrderBook, Hotspot Link). 7 Electronic trading platforms
geared to the non-disclosed inter-dealer market (eg EBS Market, Hotspot FX ECN, Reuters (Refinitiv) Matching).
Sources: BIS Triennial Central Bank Survey; authors’ calculations.

Figure 1. Spot FX Volume Distribution. Original source: BIS Quarterly Review, International banking and financial market
developments,42Dec 2019. BIS Quarterly Review, December 2019

4.1 Types of quotes Example

LPs provide two types of quotes: If quotes in Table 1 are sweepable quotes, and the client wants to
buy 3M of this pair, they can hit both the 1M quote @ 1.2610 and
1. Sweepable Sweepable quotes allow you to hit multiple 2M quote @1.2615 at the same time. If these are Full Amount
price levels or multiple quotes at the same price level. They quotes, client needs to hit the 5M quote @ 1.2620 and request
are used typically by discretionary traders who want to only 3M to be filled against this quote.
build up (or reduce) large positions in a short period of
time. LPs quote a wider spread because they are less sure Bid Size Bid Ask Ask Size
of the risk they are taking on. 1,000,000 1.2605 1.2610 1,000,00
2. Full Amount When receiving full amount quotes, a client 2,000,000 1.2600 1.2615 2,000,00
is limited to trading against a single quote at any given time. 5,000,000 1.2595 1.2620 5,000,00
Large systematic firms may favor the smaller quote sizes 10,000,000 1.2590 1.2625 10,000,00
and tighter spreads afforded by full-amount quotes since
Table 1. Sample Quotes for EUR/USD
they typically execute over a longer time horizon using
VWAP/TWAP type algorithms.

In addition, when going through an aggregator (for example, 4.2 Last Look
Bloomberg): Liquidity providers typically provide a non-binding quote and
decide whether to accept an order against this quote if and when
1. Disclosed quotes Both sender and receiver know identity they receive an order. This mechanism is called last look and
of each other. Sender can customize their quotes based on allows the LP to perform risk and credit checks. It also allows
receiver, and receiver can choose whether to trade against them to accept only those orders which are profitable for them.
a particular sender. Systematic firms can expect to receive In addition, they can hold an order for a short time before making
tight spreads on disclosed quotes because this scenario is the decision. As one would expect, in many cases, this is not
similar to maintaining direct connectivity to underlying advantageous for the party sending the order because if the LP
LP’s in the aggregator. rejects, it may not be possible to refill the order at the same price
2. Non-Disclosed/Anonymous quotes Sender and receiver from another LP.
do not know the identity of each other. This will help the re-
ceiver mask their trading activity. However, because sender
does not know the identity of the receiver, they cannot cus-
tomize the quotes for each receiver. For systematic firms,
this could lead to higher spreads.
 Execution in Spot FX — 3/5

4.3 Mark-Outs Exposure to ECNs and primary markets can help during mar-
Liquidity providers are constantly monitoring the profitability of ket dislocations, place limit orders, and generate alpha signals.
each client. One tool they use in their analysis is known as a However, trading on ECNs and primary markets carries the risk
mark-out. A mark-out plot looks at the price an order is executed of higher market impact, as any action on these markets is imme-
and market mid-price over many intervals following the execution. diately visible to everyone.
Figure 2 illustrates three different scenarios. The red line shows
the LP losing the spread (captured at trade time) rapidly. LPs will The Role of ECNs/Primary Markets in an FX Aggregator
typically mark this flow as bad for them and start quoting higher When a firm trades with a bank, the bank typically holds this
spreads for clients showing this profile to account for the rapid position for some time (i.e., has a long holding period) while
decay. Such a scenario can arise when a client has many LPs in a waiting for some offsetting flow from another client. This pro-
pool (indicating adverse selection) or uses high-frequency alpha vides for a cushion in terms of market impact. Conversely, when
signals as part of their trading. Both scenarios are bad for the LP. trading directly on ECNs or primary markets, every action on the
On the other hand, the client showing the green profile where the order book is published to all the participants. Therefore each
LP retains the spread captured at trade, will typically receive the action carries some amount of market impact with it. At the same
tightest spreads from LP. time, primary markets and ECNs are a large source of liquidity,
especially during times of high volatility in markets, when banks
might pull back their quotes. They also provide an opportunity to
 post limit orders and save on spread costs.
When choosing the counterparties to trade against, it is tempt-

ing to include as many as you can. However, this can be counter-
6SUHDG

productive to optimal execution.



Adverse Selection


With many LPs in a pool, there is a chance that one of them

 has failed to update their quotes fast enough. When a client
executes against this quote, the LP is in trouble because the market
       

+ROGLQJ3HULRG
has already moved away. As a result, the LP is sitting on an
%DG *RRG $FFHSWDEOH
instantaneous loss. If this happens consistently, the LP will start
widening their quotes and start hedging this flow immediately.
Figure 2. Markout (% of Spread retained vs Time) from LP’s
This causes indirect market impact and is undesirable for both LP
perspective.
and the client. This scenario is most likely to occur when there
are many LPs in a pool and when some LPs have a considerably
different pricing model compared to the majority.
5. An FX Aggregator The Number of Counterparties in the Aggregator
In the previous sections, we introduced the challenges posed by It is often the case that the quality of counterparties in the aggre-
systematically trading FX markets. Nuances such as last look, gator is more important than the number of LPs in the pool. As
skew and mark-out analysis add complexity to the problem of the number of LPs increases:
optimal execution.
1. Each LP receives less flow
One way to overcome some of these challenges is through
the construction of a carefully assembled FX aggregator. This 2. Adverse selection kicks in, especially with the LPs receiv-
is a system (often with associated algorithms and front end) that ing less than approximately 5% of the flow. These LPs
draws in quotes from multiple sources and venues to streamline were chosen to trade with because they could not get away
execution. during high market movement. When they look at these
An FX aggregator allows you to connect to multiple venues, trades, they will then immediately offload the position and
aggregate quotes, and view a consolidated book from which you create market impact resulting in a losing scenario for both
can choose the best quote to trade against. Below we highlight sides.
the considerations needed when constructing an FX aggregator 3. More of the flow is being directed to LPs because of their
such as the number and nature of participants in the pool along pricing inaccuracy, and this results in higher rejection rates
with their hedging styles.
By our observation, broad industry consensus is that an optimal
Who Should Be Part of the Pool? pool contains approximately 8 LPs per currency pair. This en-
It is preferable to choose counterparties with similar pricing and sures that all LPs are getting a reasonable amount of flow while
hedging models as it helps ensure that you minimize market reducing the chances for adverse selection among LPs.
impact from your execution. If one of the LPs has an aggressive
hedging model and is offloading risk much faster than others, this
information will show up on the mark-out profile, and soon other
counterparties will start quoting higher spreads.
 Execution in Spot FX — 4/5

5.1 Analyzing Counterparties in an Aggregator

It is crucial to constantly monitor the flow going to each counter-
party in your aggregator and engage with them to understand how
each manages the order flow. Firms typically look at metrics such
as order fill rates, % of flow going to each LP, last look reject
times, and cost of each reject. In addition, mark-out profiles can
show how the counterparty is managing the flow. A mark-out with
an increasing slope after execution could mean the counterparty
is immediately hedging the flow, thereby causing market impact.
A mark-out with a decreasing slope could mean the counterparty
is leading the market and is providing aggressive offers when the
market is about to go up (and vice versa).

6. Conclusion and Outlook

The first part of this paper described the challenges of trading OTC
FX, such as fragmentation, lack of standardization, and nuances
around pricing. The second part explored how the development
of an FX aggregator can alleviate some of these issues. We have
highlighted some of the considerations and trade-offs needed in
constructing such a tool: 8-10 liquidity providers per currency
pair, ideally with a degree of homogeneity of pricing and hedging
models, provides an appropriate mix. Ongoing analysis of the
flow going to these counterparties is always advisable.
 Execution in Spot FX — 5/5

Legal Disclaimer
THIS DOCUMENT IS NOT A PRIVATE OFFERING MEMORANDUM AND DOES NOT CONSTITUTE AN OFFER TO SELL, NOR IS IT A SOLICITATION OF
AN OFFER TO BUY, ANY SECURITY. THE VIEWS EXPRESSED HEREIN ARE EXCLUSIVELY THOSE OF THE AUTHORS AND DO NOT NECESSARILY
REPRESENT THE VIEWS OF GRAHAM CAPITAL MANAGEMENT. THE INFORMATION CONTAINED HEREIN IS NOT INTENDED TO PROVIDE
ACCOUNTING, LEGAL, OR TAX ADVICE AND SHOULD NOT BE RELIED ON FOR INVESTMENT DECISION MAKING.