David Mechner
David Mechner


The award winning quantitative technology provider that strives to connect traders to markets in a more transparent and intelligent manner

Pragma specializes in multi-asset class algorithmic trading solutions. The firm prides itself on a business model that does not conflict with clients’ which helps to create more sustainable, value added and long-term partnerships. We asked David Mechner, CEO of this dynamic and fast growing firm to tell us more about what his team has been doing in the FX space.

David, please tell us a little about your background and how you came to launch Pragma? 

I’ve always been fascinated by systems where complicated behavior arises out of (relatively) simple rules or components. In high school, I became fascinated with the Asian strategy game of Go, and became a serious player. After graduating high school I moved to Japan as a live-in disciple of a Go master and competed in the professional promotion system there. 

After a year and a half, I found my life goals shifting back to more traditional ambitions for an American teenager, and returned to college at NYU where I studied computer science and focused on Artificial Intelligence. 

As an undergrad I started a project to develop a Go playing AI program (sadly less successfully than Deep Mind’s recent efforts with AlphaGo). I then entered a neuroscience PhD program, also at NYU. After four years, during my dissertation research, I dropped out to join a small hedge fund that was started by friends in 2000. 

One of my main projects there was building automated trading software to execute equity strategies, and I became excited by the potential of building software that was clearly so necessary and beneficial to the industry. In 2003, I started Pragma with the goal of offering automated trading software to institutions – and though we’ve grown a lot since then, we still have the same vision and focus we did 17 years ago.

Our market microstructure knowledge and technology know-how create a powerful combination

What products and services does Pragma offer and who are you providing these for? 

Pragma is an independent quantitative trading technology provider that specializes in algorithmic trading solutions. We offer a fully managed, hosted, algorithmic trading platform for equities, futures and FX that allows clients to leverage the benefits of an in-house system, while outsourcing the maintenance and development to a specialist. 

Pragma has a unique niche with sell-side institutions across global banks and broker-dealers that want to offer high performing execution algorithms to their institutional and corporate clients. The business offers deep customization and has the ability to integrate with our clients’ own unique liquidity pool – through their in-house aggregator or a third-party – allowing banks to offer their clients a truly private label offering. 

Pragma provides the full hosting and management of the platform, the research and development behind the algorithms, ongoing maintenance of the algorithms, and first-level support directly to our clients. We also offer a series of additional tools to support our execution algos, such as real-time monitoring through our Algorithmic Management System (AMS), Panorama, our historical Transaction Cost Analysis (TCA) portal, TradeReports, and our post-trade TCA snapshot, TradeRecap. These powerful tools allow Pragma’s clients to have an experience that is equal to, if not better than if they built everything in-house. 

For the buy-side, we partner with quantitative hedge funds to customize execution algorithms across a unique liquidity pool that we host and manage. Our combination of market microstructure knowledge and technology know-how for hosting and building trading software create a powerful combination which generates significant economic value for our partners. 

In short, what Pragma offers to both buy and sell side clients is very much a partnership. As Pragma is independent and does not act as a broker or a CCP, has no principal trading operations, nor is associated with any source of liquidity, its interests are uniquely aligned with our clients.

What Pragma offers to both buy and sell side clients is very much a partnership

How has your business fared over the past year or so? What have been the key drivers of growth? 

Pragma has benefitted from the continuing trend towards more electronic trading in the FX markets, and in particular the growth of algorithmic trading. About six years ago, we entered the FX market, partnering with an existing client of ours. We launched Pragma360 FX, a unique service that enables banks to bring online market-leading execution algos for their internal traders and clients in a way that is much faster and less expensive than if they tried to build them in-house.
We’ve seen strong growth as more banks meet the demands of their corporate and institutional clients by offering them an algorithmic trading platform. As they and their clients become comfortable trading algorithmically, our overall volumes continue to grow – especially given the recent market volatility.

Why are increasing numbers of trading firms choosing to outsource their FX algorithmic trading requirements? 

We see two main reasons why firms are choosing to outsource their FX algo trading technology. The first is that demand amongst institutional and corporate traders to access FX algos continues to grow. 

Some institutional and corporate clients are going so far as to demand that for any of their trading counterparties to be relevant, they must have execution solutions across all fronts in order to do business. This means banks need to offer high quality FX research, high-touch trading coverage, streaming and RFQ, e-FX plus execution algorithms. As such, more and more banks will wish to offer an algo solution.
The second reason more firms choose to outsource their FX algo solution is that algorithmic trading is a very specialized service. To build a high quality, institutional offering requires a strong investment in time, personnel, money and internal resources as well as institutional expertise that is not easy to develop or acquire. There is also very high level of project risk, and time to market can take years given all of the infrastructure requirements and market microstructure knowledge needed. 

Outsourcing to a quantitative trading specialist like Pragma can significantly reduce the cost, time to market, and overall project risk. With demand continuing to increase for execution algos, a speedy time to market with a proven product can be quite compelling. 

Even clients with the resources and know-how to build execution algos themselves often prefer to focus their internal resources on other projects and initiatives that are more proprietary, or where there is no adequate vendor solution.

Pragma has benefitted from the continuing trend towards more electronic trading in the FX markets

What factors make FX such a good fit for algorithmic trading? 

The factors that make any market prime for algorithmic trading are fast electronic markets, fragmentation, small trade sizes and streaming real-time market data. Each of those factors multiplies the challenge and complexity of achieving best execution and increases the value that execution algorithms can deliver. 

Spot FX meets all these criteria and is therefore an excellent market for algorithmic trading. Using execution algos for spot FX has proven effective enough that traders are now also using them for NDFs, which Pragma rolled out a few years ago.

As more FX trading firms become comfortable utilizing FX algos, what trends are you seeing with respect to customized execution toolsets? 

We’ve been providing execution algos since 2003, which gives us so much breadth in how our algos can be customized. As many FX traders have been using execution algos for several years now and through a variety of market conditions, they are increasingly interested in customizing certain algorithms to meet their trading needs and styles more specifically. 

The most common form of customization we see is around routing and liquidity - such as including or excluding certain venues and which venues are used for posting or taking, or incorporating custom liquidity pools. 

An example of algo behavior customization requests we see is to adapt the rate of trading, like speeding up or slowing down when the market crosses through certain price benchmarks such as arrival price, or the trailing TWAP price.

Our interests are uniquely aligned with our clients

Most large asset managers are now using FX algos and can clearly see the benefits of doing so but corporates are taking longer to come aboard. Is the value proposition as strong for them? 

The value proposition for using FX algos is as strong, if not stronger for corporates than money managers. The reason is that corporates have zero alpha in their trades – trading only to hedge or fund a position. When alpha is low, or non-existent, spreading a trade out over time will, on average get the trader a better price. This is where algorithms like TWAP/VWAP or “Float” style strategies work well for traders to reduce costs as they can trade large orders over several hours. 

In short, trading algorithmically saves money. Additionally, execution algorithms also provide increased anonymity on orders, and since corporates can have large orders, the ability to access liquidity in the market without the corporate being identified helps reduce market impact.

Are there any lessons that FX can learn about algorithmic trading in other markets that would increase adoption rates and help to build further confidence in it? 

Market structure experts commonly refer to the trend towards algorithmic trading in FX as the “Equitization” of the FX markets. The reason is that the global equities market, particularly in the US, went from mainly high-touch to nearly 80% algorithmic in only a few years. 

The equity markets have also seen a proliferation of TCA usage coupled with increased transparency to ensure traders achieve the best possible prices in the marketplace. As FX traders continue to use, and become more comfortable with execution algos, we anticipate the percentage of order flow they handle algorithmically to continue to rise.

NDFs is likely be at the center of the growth story in FX over the coming years. Is this market suitable for algorithms in the same way as spot? 

Up until a couple years ago, NDFs were not really suitable for algorithmic trading for two reasons: first, banks and non-bank liquidity providers had been unwilling to provide dealable prices for NDFs on ECNs and inter-dealer markets. Second, banks and non-bank LPs did not have the ability to stream prices electronically. 

Both are prerequisites for algorithmic trading of NDFs. As more banks have begun streaming NDF prices and more venues support NDFs, algorithmic trading of NDFs has become increasingly effective. As such, looking across the rest of 2020 and into 2021, NDFs will certainly be a center of growth for algorithmic trading.

Outsourcing to a quantitative trading specialist like Pragma can significantly reduce the cost, time to market, and overall project risk

Pragma has a dedicated team of PhD practitioners focused on understanding market structure and its effect on trading results. In what ways has that enabled you to leverage this expertise to build better products? 

One of the main factors that separates Pragma from other vendors, as well as many firms that build their own execution algos, is the depth of our research focus on market microstructure to improve execution quality. 

As an example, let’s take something as basic as a TWAP algorithm. Most technology providers can build a basic “egg-timer” that simply crosses the spread after each defined time interval. 

But when you apply market structure expertise, you start to factor in decisions such as which venues to post at, which to take from, how to best leverage firm pricing or last look, and even time-of-day effects. 

We also understand how to modify a TWAP algo based on daylight savings, when NY and London have a four-hour time difference instead of the normal five hours. These examples demonstrate the granular level our team goes to learn in order to improve execution quality even for an algorithm as simple as a TWAP.  

The chaos surrounding Covid-19 has tested the resilience of many FX and technology providers. How has Pragma been coping with the increased demand?

Fortunately for our employees and our clients we were well prepared for working from home, which Covid-19 naturally required. Several years ago, we had to work remotely as sections of Manhattan lost power as a result of Hurricane Sandy and we couldn’t get to our office for a few days. 

Obviously, a few days is different than what will probably turn out to be several months with Covid-19. But since that experience, we’ve tested our remote capabilities annually, which has really paid off for us. The team and our systems have worked great, we handled huge trading volumes in March and April while maintaining our high level of service for our clients, as 100% of Pragma staff members worked from home.

We are two years into a project to build a new micro-trading engine the core of an algorithmic trading system using cutting-edge AI tools

Has an environment of increased volatility and difficult liquidity conditions played into the strengths of algorithmic trading? 

Algorithms are effective in a wide range of market conditions. The most effective algorithms will trade over an aggregate pool of liquidity. When liquidity is difficult, algorithms can be quite effective in sourcing the best price. 

In addition, during periods of heightened volatility, spreads typically widen. When this happens, algorithms allow the trader to post and capture some of that widened spread, rather than trading at the far touch of the market. 

Thinking about it a different way, when volatility is high and risk is expensive, breaking orders into small pieces spread out over a longer period can make the job easier for the providers of liquidity, and thus cheaper for the consumers.

Where is the next round of innovation in algorithmic FX trading likely to be focused? 

We expect that NDF algorithms – which we launched a few years ago but others are only getting started in earnest – will continue to grow faster than spot. We also see further use of artificial intelligence and machine learning techniques to further improve trading performance.

How is Pragma exploring AI and Machine Learning tools and building this into your product suite? 

We are two years into a project to build a new micro-trading engine – the core of an algorithmic trading system – using cutting-edge AI tools. We began in US equities, and are now six months into production – we are already seeing excellent results. 

We are now in the process of applying the principles and techniques we developed in that effort to FX. The challenges are quite different – because of the different quality and quantity of data available in the US equity market verses FX, and the disclosed or only semi-anonymous nature of most FX liquidity. 

But many of the core principles are applicable, and we’re optimistic we’ll be able to achieve benefits for our customers and further raise the bar for execution algorithms.

How much global growth potential do you see for algorithmic FX trading over the next few years, and how can Pragma maximize the opportunities? 

Market structure surveys, such as those from Greenwich and Aite Group towards the end of 2019, have algorithmic trading growth relatively steady at about 15% of overall spot market trading. For the real money and corporate traders that use algorithms, usage is at about 40%. This demonstrates that there are still many market participants that do not use algorithms or use it for a small percentage of their order flow. 

Our goal is to continue to increase the functionality and the execution quality of our algorithms while providing the traders with the control and transparency they need to do their jobs even better. This means we not only improve the algorithms but also the trade support tools like TCA and our algorithmic management system, Panorama. 

Ultimately, trading algorithmically is cheaper than trading manually, and the competitive and regulatory pressures to achieve best execution create unstoppable momentum. It may take years to play out, but we think the end-point of significantly higher algorithmic usage is inevitable.