PhaseCapital’s CEP Strategy

Buy-side firm PhaseCapital leverages Lime Brokerage for execution and low latency market data aggregation, integrated with StreamBase's complex event processing platform. By Emily Fraser

For high-frequency trading firms there can be no compromise when it comes to latency.

PhaseCapital, a Boston, Mass.-based quantitative investment advisor firm trading the S&P 500, chose Lime Brokerage as its execution broker for its low latency disaggregated market data feeds, and is benefiting from the broker's recent integration with StreamBase's complex event processing (CEP) technology. CEO Eric Pritchett, one of the firm's founders, talks to Waters about his firm's relationship with Lime and how it is using CEP for algorithmic trading, smart order routing and market data management, as well as to provide data to its in-house risk management systems.

Waters: What is your relationship to Lime Brokerage?

Eric Pritchett, CEO PhaseCapital: They are our primary execution broker. They have a technology platform that they offer to their execution clients. Our prime broker is Goldman Sachs and they offer us a lot of good services but in terms of really low latency, disaggregated market data feeds, and an ability to trade in those disaggregated destinations at low cost and with low latency, Lime has a pretty unique offering for the brokerage space.

Waters: Tell us about your execution platform.

Pritchett: Our primary relationship is with Goldman Sachs and our execution account is actually a joint venture arrangement between Goldman Sachs and Lime Brokerage. We have upstairs services through Goldman Sachs and trade execution services through Lime-I believe we're one of less than a dozen firms that have such a relationship with those two firms.

Lime Brokerage recognized years back that there was going to be a gap in the market for a broker that was very focused on technology and low-touch, high-frequency automated electronic trading that would be very low cost. Traditionally there have been other quantitatively-based high-frequency trading firms like ours that have gone out and opened their own proprietary broker-dealer to be able to effectively implement these trading strategies and to get the kind of broker-dealer economics that you would get from having your own broker-dealer but you would not get from a mainstream prime broker.

In the past you would have paid commissions to your prime broker for them to fly you around to conferences and get you meetings with CEOs and do the things that fundamental analysts do-read their research reports etc. That was the basis for how they earned their commissions. I think what Lime realized very early on is that there is a segment in the market-the quantitative high-frequency traders that are looking for broker-dealer economics without having to open their own broker-dealer, and they would rather use their commission flow to pay someone for useful technological resources that they would otherwise have to build. They don't want to pay for glossy research reports about industries and things like that because it's not what they do.

The core of their electronic trading platform is the open protocol FIX 4.2 and they have done some good things on their side to optimize what is otherwise an open standard, to compress it and make it very efficient for high-frequency trading. They don't provide that out to us but they give us the peculiarities of their implementation of FIX and then on our side we've implemented QuickFIXJ, which is a Java version of the QuickFIX engine. We've implemented to version 4.2, we certify our protocol against their protocol and then once testing is done the two systems can talk to each other. And the result of that is called a FIX session and that FIX session is kept in memory and managed in StreamBase.

Waters: How is StreamBase complementing this relationship with Lime?

Pritchett: When you go into Lime to integrate with them, while they do have a strong API that is available against their market data and their FIX engine platform, there is a pretty significant integration task that remains for the firm that's integrating into those capabilities. When you bring StreamBase in you are able to use adaptors that are purpose-built against the capabilities that you're trying to leverage within the Lime technology arena and you're able to put the developer team into a very strong end user development paradigm with their development tools.

StreamBase's core technology interfaces with Lime-maintaining the low latency that we're getting from Lime. It holds the valuable data in memory and allows us to develop and integrate our trading strategies up against the marketplace. It also allows us to implement our order routing in the StreamBase platform as well as our feed handling and trade session management.

Waters: Does your firm use the StreamBase platform out-of-the-box or have you tweaked it?

Pritchett: Our employees include people with PhDs in hard sciences, Masters degrees in engineering fields as well as investment professionals with a heavy technical orientation. Our team has a very hands-on, down-to-the-bare-metal approach to engineering our systems and so while StreamBase puts very strong templates and a very strong functional base into the hands of our developers, we don't expect any technology vendor to provide something out-of-the-box that is going to do everything we need it to do. There are CEP solutions out there that may sell themselves as a little bit more turnkey or out-of-the-box, but in those cases the abstractions make us a little uncomfortable. For every bit of turnkey that you get, you lose some control. StreamBase's platform and end user paradigm makes our developers comfortable integrating the technology as a core part of our own trading system.

Our developers connect via their workstations into our Linux cluster and our Linux supercomputing back end. The team that has integrated with StreamBase is a hardcore engineering team-but that's not to say that StreamBase would be difficult for a more traditional IT group to work with.

It is critical to understand and focus on your value-add, and our unique value-add is in our signaling, in our opportunity identification, in our position sizing, in our risk process and in realizing the alpha opportunities that we are trying to harvest through high-frequency trading. It's not value-add to have our team trying to do the sort of engineering in memory that StreamBase and Lime Brokerage have done-these are things that you would really rather buy than build if you feel comfortable with what's under the hood.

Waters: According to StreamBase officials they are seeing the most interest in their platform for market data management, SOR, algorithmic trading and foreign exchange (FX). How are you using StreamBase?

Pritchett: We're using StreamBase for all of those areas. Our initial universe that we are trading is the S&P 500. We have CEP from StreamBase in the execution management area, which includes how we implement our proprietary order routing. We have it in between us and our two large market data providers, Thompson Reuters and Lime Brokerage's market data platform Citrius. We are also beginning to implement it for providing data to some of our in-house proprietary risk analytics and real time P&L calculations. Going forward we expect that CEP will play a similar role in terms of execution management, data management and provisioning data and information to risk and monitoring applications as we expand to different asset classes like currencies and futures.

FX is a very natural application for StreamBase because FX doesn't really have a true exchange place. If you want to trade FX, you have to get individual data feeds from large dealer banks and aggregate those different transaction opportunity sets. So its similar to how equity trading has become disaggregated. Currency trading has always been disaggregated.