January 1, 2020 - A guest article by Karl Elsammak, Head of Trading and Founder of Kammas Trading.
Carl Elsammak, Founder, Kammas Trading
Much of my career has been spent trading G10, NDF and exotic currency pairs for both Buy and Sell side participants with an emphasis on dealing / market making, risk mitigation / hedging, execution, and systematic trading. In the early days of retail FX I established my company Kammas Trading to provide brokers with the dealing desk / market making services and liquidity management they found difficult to assemble locally. We have gone on to manage dozens of broker’s client flow totalling well over a trillion dollars. During this time, it has become clear to me that several misunderstandings recur on the topic of B-Book management.
The Benefits and Risks of Running a B-Book
B-Book management is the pricing and acceptance of a client’s trade request onto a broker’s own books. In short, the client’s trade is not immediately hedged. Immediate hedging is defined as A-Book. Brokers who decide to run a B-Book do so primarily because the profit potential per unit traded is consistently higher when responsibly managed, than an A-Book can provide. In addition, a B-Book broker has significantly more control over pricing and fill characteristics. As a result, their clients receive improved market conditions and regularly report greater customer satisfaction when compared to A-Book brokers. The caveat being that additional risk is associated with B-Booking activity. Holding the resulting positions and exposure from tens of thousands of client trades daily can result in significant losses if done poorly by an inexperienced dealing team. All B-Book brokers manage a certain % of flow via A-Book techniques.
By A-Booking a broker does not eliminate all risks just by running a matched book (all client trades hedged at liquidity providers). They remain subject to gap risk; a situation where multiple client’s accounts result in a negative balance as the result of a large gap move. These negative balances are difficult if not impossible to collect on however; hedged trades at the liquidity providers will reflect the full loss to the broker. The most spectacular of these events occurred on January 15, 2015 when the Swiss National Bank removed its currency’s peg to the Euro without warning causing huge currency moves that resulted in some of the larger brokers losing hundreds of millions of dollars in an instant.
Much marketing has been published (by A-Book brokers) trying to convince retail clients that B-Book brokers are “trading against them” or involved in some nefarious activity designed to steal from them. Overwhelmingly this is simply not the case. Regulators have all but eliminated the early brokers who engaged in these practices. Further, risks to retail clients normally happen at the point of execution. A-Book brokers have just as much ability to manipulate executions and fills as their B-Book counterparts. I am not saying there are shenanigans occurring at A-Book brokers, just that their suggestions it happens with greater frequency at B-Book brokers are unfounded.
My intention for this article is to help illuminate some of these issues and allay misconceptions by explaining, in broad terms, the inner workings of a broker with a predominantly B-Book approach.
Professionals new to the industry or focused in areas outside of trading should find this useful in establishing a more comprehensive understanding of the complexities of running these businesses.
Retail clients should take solace in the fact that most FX / CFD brokers are controlled by industry professionals who run trustworthy, respectable businesses designed to provide you with the best access to these markets. I hope after reading this article you gain a greater appreciation for the various considerations necessary to do so.
Differences between Institutional and Retail market making
As we consider how to manage trade flow within a retail FX broker, let us first identify some of the major differences between this environment and an institutional market maker such as a bank. These contrasts include ticket size, frequency, and revenue expectations. The clients of institutional market makers include Prime of Prime brokers, corporations, hedge funds, asset managers, retail brokers, smaller banks, etc. As you can imagine their average ticket is significantly larger than that of a retail client. In addition, the number of live active accounts within a retail broker are exponentially larger than at a bank. Finally, a bank’s risk tends to come in large chunks from professional traders. This results in more binary outcomes with a negative bias. Therefore, duration and revenue capture expectations are reduced within a bank when compared to a retail broker.
Price discovery, Spread, Skew, Fill time, Size
Market makers, regardless of their designation, have many tools available to help them manage their business interests through the study of market microstructure, the study of how efficiently a market operates within its given environment and ruleset. The foreign exchange market, like all over the counter (OTC) markets, is highly fragmented creating innumerable trading environments designed for a specific use or client making this study particularly important.
Major considerations for pricing include bid / offer spread, skew, refresh rate, and size. As trade requests arrive, the consideration of expanded fill latency can be added in times of enhanced price volatility. Each of these will be explored in the coming paragraphs.
Often the best method for accomplishing price discovery in a fragmented OTC market like FX is to determine separately the Bid and Offer prices from a variety of other liquidity providers (LP) whose price is trusted at that moment. While it is easy to identify top of book (TOB) or the highest bid and lowest offer available, this is not the best method for determining where a market is at any given point. Each LP will be applying their own pricing skew causing TOB prices to be the most exaggerated and bespoke examples at any given time. While many factors can and should be considered, an easy method is to determine the price-weighted average where a specific amount can be executed… i.e., $3 million (this amount is unique to each entity and will depend on the depth of trading lines available).
(diagram/Illustration of the book here?)
For example, if $1 million is bid at 55.3 by LP A, $1 million at 55.4 by LP B and $1 million at 55.5 by LP C then the bid for price discovery purposes of $3 million is 55.4. Do the same weighted average calculation for the offer to determine the market price at that moment as represented by your LPs. As a side note, take notice when there is significantly more size on one side of the spread within the same price range as the other side (bid vs. offer) (illustration?)since this may highlight an impending liquidity vacuum due to large orders at your LPs.
Next, the market maker should determine what neutral spread they want to show their clients. The neutral spread is the price discovery determination from the above paragraph adjusted (normally widened) to the spread range to be shown to its clients. This spread is governed by the business terms, commissions, and competitive expectations each broker markets to their clients, and should be represented internally as a range rather than a specific number. Clearly the neutral spread should be wider than pricing available from their institutional providers so a simple widening algorithm will do the trick.
Now that we have a neutral price, we must consider our unhedged exposure in any given pair or currency that is part of that pair. By applying a skew, we are encouraging activity on one side of our spread while discouraging activity on the other. If a broker is long Euro and wants to reduce exposure, they will skew their prices lower on all EUR/XXX pairs encouraging activity on the offer. (This assumes that XXX is not also a long position that needs to be reduced.)
Generally, a broker will want to fulfill trade requests as quickly as possible. However, during volatile markets price discovery can be compromised and trade requests should be held for longer periods to ensure a fair pricing determination is made for both parties. To be clear fill times are measured in milliseconds or 1/1000th of a second so delaying a fill by 50 to 500 milliseconds will be indeterminable. Volatility should also be the trigger to allow for fill time adjustment and spread expansion / widening beyond the neutral range.
Finally, a broker can also limit the size available separately on the bid and offer as a further form of protection.
For many pairs including less active crosses, it is often preferable to determine pricing by triangulation vs a common 3rd currency. By doing this a broker can better control the various considerations described above by concentrating on higher volume components. These currency components have a greater likelihood of regularly accumulating significant directional exposure making their pricing specificity more important.
Maybe another diagram?
Consolidation of risk into silos & Internal hedging
It is essential that a broker determines an accurate, comprehensive position for all products in real-time. Some brokers simply keep their positions in every pair or instrument traded. However, this is an inefficient method that results in unnecessary hedge transactions and reduced revenue. A preferred method is to establish risk silos where the common currency from all applicable pairs is consolidated. By doing so, a broker will naturally capture a significantly larger portion of the quoted spread while simultaneously reducing the overall internal exposure efficiently. Remaining exposures can be seen and therefore managed with significantly greater clarity via any pair.
Creating risk limits – pragmatic tools
Risk limits need to be established and adhered to. There are many complicated methods for measuring risk including value at risk (VAR) analysis, underperformance risk, journey risk, and capital-at-loss risk just to name a few. However, the real-time tools need to be pragmatic in their design and implementation. I find that simple position limits work best. These limits should be measured by currency and overall position. By currency, individual considerations should include volatility and transaction frequency with higher volatility and lower trade frequency pairs receiving lower position limits. Overall exposure is not the sum of the various max positions, but rather a total open position that represents acceptable loss considerations given 2 or 3 standard deviation moves. To expand on this, the overall position number should take correlation into consideration. The construction of a simple matrix will go a long way to accomplishing this task. By doing so, a net long Eur position of 10 million and a short CHF of 8 million does not contribute USD 18 million to the overall exposure. Instead, the overall position exposure effect would be 4 or 5 million due to their normally high correlation and offsetting position. (Numbers are for illustrative purposes.)
Business intelligence (BI) has become an integral part of all successful companies regardless of industry; often devoting entire departments to the task. All major brokers use these measures to better understand their business. These are among the most valuable tools available as they provide in addition to effective planning tools, an early alert to problems as performance strays from the norm. Volume, revenue, pip capture, average trade size, average trade duration, each calculated by pair and total as well as broken down into various time slices and by client definitions are all examples of necessary measurements. Regularly studying these not only highlights issues but dramatically helps to identify their origins quickly. It also gives management a sense of comfort and tremendous insight by understanding their business with greater granularity. While these can be built internally, some very sophisticated services from companies such as FairXchange and Centroid offer a quick path to these essential tools.
All responsible brokers have an active hedging process, and a need for quality liquidity from a variety of providers is incredibly important. Many brokers feel that by simply adding as much liquidity and as many providers as possible they will achieve the best results.
However, this is not the case.
Very simply put, all liquidity is not alike, and care must be taken to define, negotiate and select lines with appropriate size, spread and fill characteristics. Last look pricing is acceptable if the LP is acting in good faith and may be preferred to a no last look feed which tends to have wider spreads. Likewise, brokers must be careful not to abuse LPs with toxic flow like arbitrage or frequent lumpy orders.
Institutional liquidity providers are for profit organizations and will always place their interests above yours. Whether it be a bank, HFT, larger broker, or hedge fund these firms have become highly sophisticated, adjusting their pricing and fill characteristics to each client. These characteristics change dynamically based on risk factors accumulated by algorithms over time. However, while there certainly is a strong quantitative factor, each LP will also adjust manually. Therefore, qualitative considerations are often equally if not more important. As a result, relationships still matter! Pick up the phone and chat regularly with each LP. Provide them with feedback; they love it! I often maintain statistics by LP and pair on TOB, volume, average spread, dispersion, fill time, rejection rates, etc. Knowing where they stand against the rest of the pack is great information…just, don’t reveal who the other LPs are. Maintain a variety of LPs with different designations but be sure there is enough business to keep them all happy. All of this builds trust and will improve spread and fill performance as your LPs work to keep you happy.
Respect the client
Let us agree on a few basic tenets. A client’s deposit belongs to the client not the broker (remember it is a liability on the broker’s books). Brokers make revenue primarily from transactions with an expectation of earning roughly the spread, on average over time. A valuable client is one who actively and regularly transacts, regardless of their ultimate P&L. Clients are expensive to obtain. Therefore, each broker must treat their client base with respect by providing a fair dealing environment. Clients are not stupid (generally) and will recognize when treated poorly. Some examples include asymmetric slippage, skewing against limit orders and toward stops, blatant rejects during volatile times, excessively wide roll / financing / swap charges, unnecessary fees, etc., etc., etc., the tricks abound. While enticing and profitable, these practices are exceptionally short sighted. Broker after broker have fallen into this trap only to see their account growth slow, volumes drop, and in the worst cases regulator fines accumulate. A bad reputation is a death knell in this business. The point is: Define and communicate your offer clearly then keep to your promise. Most brokers understand this and act accordingly.
Experienced dealer’s vs Algorithm
Brokers who run their own risk need to determine if they would like a rule’s based algorithmic process that offsets exposure on an automated basis or have an experienced team actively manage this essential function. As with many of these choices I believe the answer lies somewhere in between. While its enticing to believe that your entire risk mitigation and revenue generation can be left to a program, reducing salaries and other overhead; the truth is that these programs, however sophisticated their origins, always leave a tremendous amount of profit on the table that will far exceed any cost. Simply put they are unable to adapt to changing market conditions regardless of their designer’s claims.
While we can all agree there is no substitute for true experience, even a highly accomplished dealing team working in a vacuum without the support of well-designed and maintained alert tools will have trouble producing a desirable, consistent return profile. However, a professional team that shares information, actively inputs market range expectations with defined action points, and works with a strong leader to develop tools that provide enhanced clarity to the future risk return profile of a given position at a given price and time will always outperform.
To be clear, performance is best measured as the amount of return achieved for a given amount of risk accepted. It is the risk portion that needs to be defined before an appropriate strategy can be formulated. Brokers who tend to focus solely on the return aspect with little regard for or understanding of the risk exposure quotient, often end up with catastrophic outcomes for both themselves and their clients whose deposit has been entrusted to them.
While the above article is only a cursory discussion on this topic, I hope it helps to provide some insight to the complexities of running a market making operation within a retail FX / CFD brokerage environment. Should you have any questions, comments, or would like to engage in a conversation please feel free to contact me directly, my information is below.
IS Prime Partners with Reactive Markets to Launch New Multi-Asset Trading and Risk System
ATFX partners with Tier-1 Prime Broker
OANDA Offers Cash Rebates To High-Volume Clients In The US
Raiffeisen Bank International Partners with AxeTrading and Integral to support bond trading
FXCubic appoints Richard Bartlett as Head of Sales