Getting the Right Fit : Buy-side, Sell-side and Vendor on Non-Standard ATDL Implementation


Dmitry Koltunov of Highbridge Capital Management opens up to FIXGlobal about the buy-side expectations for FIXatdlSM algos and how vendors and sell-side firms can improve their offerings.
How do you evaluate a broker’s FIXatdlSM offerings? Are you requiring or looking for them to hold to a standard?
Our preference is towards FIXatdlSM 1.1 or 1.0, since the recent iterations are able to capture more details of the algorithm. However working with any version of FIXatdlSM is already a significant improvement over a static document. Since our proprietary EMS/OMS is built around a customized implementation of FIXatdlSM, the broker FIXatdlSM can never be taken “as-is”. We also do some minimum tweaking to the layout, rules and default settings based on our trader’s preferences. This allows us to have the same look and feel for similar algos from multiple providers.
The standard has been evolving significantly, and in our view, it would be beneficial for the industry to come to more of a convergence point. Having every FIXatdlSM consumer require a different version may be detrimental to overall adoption, because it raises the maintenance cost for the vendors to keep backwards compatibility for each client. The FPL Algorithmic Trading Working Group (which developed FIXatdlSM) has now drawn a line in the sand with the latest version and is making a widespread industry push for it. FIXatdlSM 1.1 is quite comprehensive, so we would be pleased to see the community shift gears from evolving content to aiding adoption of this version. However there are constantly new types of Algorithms coming to the market and this will keep pushing the standard to grow and change. The challenge remains for the industry to work collaboratively to evolve at a measured pace while fostering widespread adoption.
How responsive are your brokers to your requests for new algorithmic products?
We have seen varied response times, which range from a few weeks to over a year, depending on the broker, algo type and level of customization. Most bulge bracket firms are responsive, but the smaller boutiques can vary from poor to great, with those that explicitly focus on execution at the forefront. Some brokers have also put together custom solutions for our traders, and those timelines vary from weeks to months.
It has been a challenge to integrate with brokers that tie their algos to a particular EMS. In one particular situation, a vendor indicated that it would take over a year to make its algo accessible through FIX. Waiting was not an option, so we ultimately worked with another vendor with similar functionality that did not require a particular EMS and could integrate through FIX and FIXatdlSM. Vendors that realize that their real product is execution logic, as opposed to a particular display or EMS, wind up being more successful. In general, we have recently seen more brokers respond to us with FIXatdlSM. This is a positive development, because a year ago, only a handful of brokers were even aware of it. In many cases, we were the first firm to ask our brokers for FIXatdlSM files, so we find it encouraging that more than half are now able to serve up the files. When we do request FIXatdlSM, the timelines vary from weeks to months, and we have seen some firms hire consultants to complete it more quickly.
Timing is a differentiator and FIXatdlSM is critical to this. In general, traders become accustomed to how certain algos behave, and the first algos to get ’test-driven’ have an advantage. As the space becomes increasingly crowded, I can see the laggards start to struggle to keep a presence.
From the buy-side perspective, how can a broker use FIXatdlSM algorithms to their advantage?
FIXatdlSM provides cost efficiencies in onboarding algos by dramatically reducing certification time. Algo providers can test their FIXatdlSM files before clients receive them, which dramatically reduces the likelihood of misspecification. Since the buy-side can process the FIXatdlSM file directly into their OMS, the potential error in translating spec documents into code is removed. Unforeseen misspecification, however, can still occur. If thisvhappens, a good implementation of FIXatdlSM will allow for on-the-fly adjustments. The buy side tester is empowered to correct mistakes during the process and continue the certification, rather than going through another round of Graphic User Interface (GUI )development. Nimble responses reduce the time period for certification from weeks to days, improving timeto-market. We have a long queue of algo requests, which allows the brokers that provide FIXatdlSM to enter our system more quickly. Another major benefit to FIXatdlSM is that it provides high-detail customization at a low cost, which empowers the algo sales team. With FIXatdlSM, the same algo can be set up with different defaults or custom dependency rules, according to the specific needs of the client. Vendors can also expose different optional fields to each client rather than the current ’one size fits all’ approach. This removes the need to have confusing generic placeholder fields (i.e. “Option 1”) that vendors use to avoid extra GUI development cycles.
The biggest win, though, is time-tomarket, as enhancements to algos can be seamlessly pushed to the clients. Depending on the amount of time needed to re-certify, new functionality can theoretically be made available within days, allowing for a heightened level of service.
What are the main issues you encounter when using FIXatdlSM algorithms, and how do you see them resolved?
A potential issue with FIXatdlSM is the strong focus on evolving the standard, without an equivalent effort to provide sufficient coding tools to facilitate adoption. The FPL Algorithmic Trading Working Group must make it easy for organizations to integrate and keep up with updates. One solution could be for the Working Group to produce Open Source software in multiple languages. A base level of functionality to render/test the screens, with hooks allowing for customization, would drastically reduce the burden of implementation for new entrants. Making the software backwards compatible would facilitate easier upgrades.
As the standard evolves, early adopters must consistently play catch-up. We started using FIXatdlSM a few years ago in its nascent stages. At that point, there were many gaps that we needed to fill with custom solutions and proprietary tags. The recent versions have addressed almost all of our needs and the functionality is now quite comprehensive. However, the recurring need to upgrade our infrastructure has slowed down our ability to incorporate newer versions.
From a technical perspective, there are two logical pieces of infrastructure that require updating with new versions.

  1. Parser: This code reads the XML and creates Data Objects that can be used programmatically. Our parser can process multiple versions of ATDL, which produces generic Data Objects that our application leverages to construct the GUI.
  2. GUI Generator: This code interprets the Data Objects and generates screens on the fly with appropriate layout, rules and validations. These screens are constructed to produce the FIX Tags to be sent by the EMS.

Earlier versions of FIXatdlSM changed quite significantly from a structural standpoint, requiringmany iterations of the Parser. For instance, the placement of Strategy Parameters shifted between Top Level and Strategy Level, which required code changes. The more recent versions come with fewer structural shifts, and more often introduce new elements/tags/attributes, which makes it easier to update the parser. The parser is a good example of something the Working Group can provide as Open Source, as it will help manage version upgrades smoothly and maintain backwards compatibility.
Once the parser is upgraded to produce the appropriate Data Objects, the GUI Generator must, occasionally, also be upgraded. Here, two changes in FIXatdlSM are important to note. First, simple structural moves (i.e. shifting Strategy Parameter locations) should require no changes and produce the same screens; a good parser implementation will abstract this from the GUI Generator. Second, whenever new elements/tags/attributes are added, new code must be created to leverage the changes. This is another place where the Working Group can create Open Source software: a library that generates algo screens and has hooks to allow for proprietary layouts, so that every EMS can have its own look and feel. The layout and general look and feel is custom to each system, but the creation of the screens, their functionality, and the FIX values they produce is fairly generic.
As long as the FPL Algorithmic Trading Working Group grows FIXatdlSM and produces new versions, code must adjust to these changes. Many of the software updates necessary for each new version of FIXatdlSM are common, and Open Source software provided by the Working Group goes a long way to ease the development burden.
Where do you see algorithmic trading expanding: e.g. new trading arenas, asset classes, etc.?
I have seen an uptake in algos that trade multiple assets together, and I think this will continue to grow. There has been increased interest in Pairs Trading, with a number of new players entering the market. I think the execution quality and adaptiveness of these algos still has room to grow, as many are in their first generation. I can also see growth in Option algos, which can help facilitate spread trades and achieve floating delta exposures, as well as the development of Basket algos to facilitate portfolio rebalancing.
In terms of asset classes, I foresee growth in algos trading Commodity Futures, where the market is currently lagging. I think there will be more focus on ETFs and potentially automating the trading of Index Swaps. I suspect we will also see algos starting to trade across asset classes. There is an appetite for trading Equities against Futures, or ADRs with an FX component to keep constant currency exposures. I think the overall industry will move towards focusing on exposure over particular asset classes.
I think that the next generation of algos will become much more adaptive, stepping away from a single execution pattern (e.g., VWAP or IS). This generation will monitor the market as a whole and contain various modes, transitioning in response to volume, volatility, and price points. We have already seen, from a few vendors, the beginning of this evolution. This will be a differentiator since it allows traders a ’best of many worlds’ scenario and will model real trading more accurately. I also see automated feedback from algos coming into focus, where some of the analyses used by the algo to make decisions is kicked back up to the traders for enhanced market color at some manageable frequency. There is already nascent motion on this as well.
How to Get the Most From FIXatdlSM : FPL Global Technical Committee
When traders do some minimal tweaking of the broker supplied FIXatdlSM, often they are effectively extending the FIXatdlSM to include certain proprietary OMS-required information. One elegant approach to this uses XML namespaces to extend the FIXatdlSM by referencing the standard schema as a component in a larger schema that include the additional information. This means that the FIXatdlSM remains completely compliant with the official schema, and any local customisation information can be carried alongside it.
While some industry participants have commented on the rapidly evolving nature of the specification, this is the trade off volunteers make when they are involved in the development of a standard. Effectively, they trade influence over the direction of a standard for the inconvenience of working with an evolving specification; however, since the release of  FIXatdlSM v1.1, which is the stable, approved version, there have been no changes to the specification.

Deutsche Bank’s Boris Goykhman talks with FIXGlobal about how DB walks the line, trying to match different vendors’ FIXatdlSM offerings to clients’ needs.
Boris Goykhman, Vice President, Product Development, Deutsche Bank
How do you evaluate a vendor’s FIXatdlSM offerings relative to other vendors or relative to your existing algorithmic products? Are you requiring / looking for them to stick to the ‘standard’?
The most important aspect of a vendor’s FIXatdlSM offering is how closely they adhere to the FPL recognized FIXatdlSM 1.0 or 1.1 protocols. The purpose of ATDL was to define a protocol that would be widely adapted and supported uniformly throughout the OMS/EMS and Broker Dealer community. In order for a vendor to have a successful FIXatdlSM offering, they must be able to support all data types, validations, control workflows, and GUI elements defined in the FIXatdlSM protocol. The expectation is that they would be able to accept our Algorithmic Ticket ATDL file, and seamlessly load it into their environment. This serves as a minimum requirement for classifying a vendor as being FIXatdlSM compliant. Additional features that a vendor may offer include Algo Ticket viewers, XML file editing environment, and an online repository for updating files.
Any vendor that does not adhere to the protocol is not classified as FIXatdlSM compliant and we would not be able to leverage existing code. A number of vendors use proprietary XML based syntax for defining broker Algorithmic tickets. Working with these vendors requires a full implementation cycle to generate code that could not be leveraged for any other vendor.
How would you characterize your clients’ expectations regarding FIXatdlSM and algorithmic trading? Is it expected, appreciated, a competitive advantage?
Clients that use a vendor that is FIXatdlSM compliant have a much better expectation of time-to-market for delivering our products to them. When utilizing an environment that allows the broker to make updates to their Algo ticket offering, the speed at which these changes can be reflected on the client’s machine is greatly increased. With FIXatdlSM we could deliver changes to our clients within 1 week of the original request. This includes requirements gathering, implementation, testing, and upgrading. Without FIXatdlSM, this process takes roughly 2 months.
As a firm, we design our Algo Ticket offering to fully utilize the features and benefits associated with FIXatdlSM. This includes control workflows for conditionally required parameters, validation of trader input, and tool tips to help our clients. Overall, these elements add to a pleasant and feature rich experience that our clients have come to expect. The competitive advantage is shared by our sales teams and vendor partners.
How do you use your FIXatdlSM offerings as a distinguishing factor among your competitors?
By offering a product that is highly functional, customizable on a per client basis, unproblematic to implement and distribute, and has a familiar look and feel, we are able to keep pace with the ever changing algo trading industry. The ability to quickly evolve, adapt, and deliver sets us apart from other brokers and vendors.
What are the main issues you and your clients encounter when using FIXatdlSM, and how would you like to see these resolved?
The main issue encountered with FIXatdlSM stems from lack of tools and enforcement of the FPL defined protocol. Every vendor is tasked with developing and integrating the three components that allow them to support FIXatdlSM. They are required to create a Parser (reads the XML filed containing the Algo Ticket implementation), Render (Generates a GUI used by the client to create an algo order), and data model for processing validations and generating a FIX order message. All of these components need to be integrated into their OMS/EMS and the end result is a unique offering from each vendor. The expectation is that the behavior across platforms will be uniform, but that is not always the case.
Additionally, FIXatdlSM is limited to defining the Algo Ticket and it is impossible for the broker to predict how it will be integrated with the EMS order entry function. Some vendors offer to include basic order parameters (OrderType, Limit, TIF, DisplayQty, etc) within the same window as the Algo ticket, but those who don’t create a less then ideal workflow for the client. In the future, FPL should include more stringent guidelines for how the Algo Ticket will appear in the OMS/EMS and provide source code for developing the Parser and Render elements.
Where do you see algorithmic trading expanding; e.g. new trading arenas, asset classes, etc.?
Over the last few years, algorithmic trading has become main stream for single stock trading. As the environment continues to evolve, it is very likely that algorithmic trading will become the norm for FX, futures, options, and fixed income as well. FIXatdlSM is primed to play a role as algo trading continues to gain traction and will very soon dominate the landscape.

Highlighting the perspectives of FIXatdlSM technology providers, FIXGlobal asked RealTick’s Stuart Breslow to offer insight on his firm’s experience of FIXatdlSM and how clients view the standard.
Stuart Breslow, CEO / CIO, RealTick
How do you see your clients evaluating FIXatdlSM offerings relative to their existing algorithmic products? Do you see it as important to stick to the ‘standard’?
As a leading vendor and early proponent of FIXatdlSM, we at RealTick believe it is important to ensure the latest and most comprehensive FIXatdlSM standards are understood and supported by industry participants. In fact our firm was involved from the start with the FIXatdlSM working group and schema, so our team has built up a detailed understanding of FIXatdlSM, from the analysis of files, through to quickly and effectively completing any changes to ensure a broker’s file functions optimally within our system for their clients. Our aim is to ensure we adhere to and support the latest schema, while maintaining the flexibility to support specific broker provided XML files, which may have additional attributes the broker requires. New releases from FIX have to be reviewed and upgraded quickly so new user features can be adopted. For the industry as a whole it makes sense to create and adopt agreed standards and iterative releases, as this means we collectively lower the opportunity cost and time to market to get new algorithms into the hands of traders. However we also want to ensure our partners can innovate and the ability to provide the latest and advanced execution capabilities to their clients is not restricted. Hence a key element in this process is having the necessary depth and breadth of experience that enables a vendor to support the FIX standard across the user base, along with the flexibility to supplement this when required to do so.
How would you characterize your clients’ expectations regarding FIXatdlSM and algorithmic trading? Is it expected, appreciated, a competitive advantage?
Back in late 2008 and through our involvement with the FIX technical committee, we saw the opportunities FIXatdlSM afforded us as a business and the benefits to clients. So we started working on being able to adopt the protocol, as well as provide internal and client (broker) facing enrichment tools that would enable us to use FIXatdlSM even more effectively. RealTick started to work closely with a number of key broker dealer partners in early 2009 to implement their equity algos using FIXatdlSM. At that time the uptake was still fairly limited across the industry, but our partners were keen to work with us and together we deployed a number of new algorithms and refined the process along the way. The experience the brokers had was very positive – FIXatdlSM allowed much of the workload around implementation, plus defining algorithmic enrichment screens and layout, to be simplified. Furthermore their new algos were getting out to their institutional clients much quicker. As a result the brokers could focus on creating new algos or customising existing ones and then deploying these to clients more efficiently. If you can see and exploit changes in a macro marketplace or region, or micro liquidity structure and then deploy a new algo potentially within days, that has to create a competitive advantage, for those providing brokerage services and traders seeking to optimize execution performance and investment strategies. For the major brokers, and certainly for RealTick, we believe FIXatdlSM is now the de facto method to support algorithmic trading.
How do you use your FIXatdlSMofferings as a distinguishing factor among your competitors?
One of the challenges vendors faced was how to fully leverage the potential of a comprehensive new protocol for clients. We wanted to ensure we and our partners had the ability to quickly and easily use FIXatdlSM. So RealTick enhanced its existing complement of web based broker dealer tools, with a suite of purpose built screens that allow brokers to upload and view their XML files, verify FIX tags and descriptions, plus attach their own branded logo to the algo enrichment ticket. Importantly if the file does not conform to the FIXatdlSM specification, those specific items and lines are highlighted to the broker, along with a brief description as to why, so the broker can quickly rectify the line and reload the file. It really becomes a quick and painless procedure, removing iterative development and testing cycles and enabling the broker to easily and proactively self manage the process of deploying new algo releases to clients, or for us to manage the entire process for the broker. Importantly for brokers and traders, via a robust Software-asa- Service deployment model these new algos are immediately visible and ready to use on the traders’desktop once tested and released.
What are the main issues you and your clients encounter when using FIXatdlSM, and how would you like to see these resolved?
FIXatdlSM was forged in the equity space and initially catered for commonly used order types. As liquidity access and markets continue to challenge investors who require enhanced execution capabilities, we will need to see FIXatdlSM broadening its reach to include complex order types and linked trading strategies, such as pairs and spreads trades. As brokers and traders look for advanced functionality to be incorporated, the collective challenge is to ensure FIXatdlSM is able to accommodate the range of new functionality, markets and capabilities in a timely manner, while still maintaining the integrity of the FIXatdlSM schema and vision as a whole.
Where do you see algorithmic trading expanding; e.g. new trading arenas, asset classes, etc.?
With market volatility and structural changes likely to continue as part of the trading landscape for some time, it is fair to say algorithmic trading will continue gaining ground and usage. Liquidity fragmentation continues to create challenges and opportunities, with the deployment of new algorithms playing a central role for investors seeking to achieve better execution. A recent TABB Group study predicts algos will soon account for 35% of buy-side order flow in the US and similar growth trends we believe hold true for Europe. In addition we see increased algorithmic trading being adopted across the Asian markets, as well as in ‘newer’ regions and markets such as South America. In fact a number of our broker partners have been adding or expanding existing algo trading capabilities for clients using RealTick in these markets this year. Other areas of growth are the adoption of ‘low touch’ algorithmic trading for linked trades, such as equity pairs and spreads trading, as well as broader algo asset expansion in options, futures and foreign exchange markets. These trends still have some way to go and as brokers continue to innovate and FIX expands its capability to accommodate new requirements, we expect the development and uptake of algos will accelerate and broaden, to the benefit of brokers, vendors and traders.