The StreamBase Event Processing Blog

Fast is Good, But Fast and Dumb is Dangerous

Last week I was on a panel at the Bloomberg Enterprise Technology Summit (@BBGlink) with Steve Goldman from CME Group (@CMEGroup), the CTO of NYSE Technologies, Bloomberg TV Markets Reporter Dominic Chu (@The_Domino) and we talked about how speed is important, but that the real innovation going on today is about making applications smarter, not just faster. The conversation was constructive and productive and was re-tweeted all over the internet. 

Then, this weekend, Barron's article "Next Danger: 'Splash Crash'" featured an interview with Progress Software's CTO Dr. John Bates. Barron's cites his view that high frequency trading poses an "extreme risk" of creating a "splash crash," his own term, which he describes as a worse than the flash crash: "trading systems could clog up, limited bandwidth could choke orders, exchanges could freeze up -- splashing across all of the affected asset classes. Pandemonium."

Bates "frets that, absent more oversight, terrorists wielding smart machines could attack the markets in an attempt to cripple our economy,"  and has indeed been fretting publicly for a while now in interviews blog posts, such as:

• "The trouble with algorithms: wild children or reckless parents?"
• "Beware the splash crash"
• "Red flags in morning, firms take warning"
• "When does a rogue trader become a scoundrel?"

It's unhelpful to focus on the fear; we should focus on solutions. Every automated application - from kiosk-based airport check-in to security in a retail store - needs surveillance. Trading is no different. This is not new. What's new is that trading occurs at speeds never before seen, and when things go wrong, it's easy to blame the bankers of nefarious activity.

So today's article in MIT Technology Review, "Faster than a Flash," was a refreshing break from provocative predictions of doom and gloom. MIT features Boston-based proprietary trading firm PhaseCapital (disclosure: PhaseCapital is a StreamBase customer). CEO Eric Pritchett described to MIT reporter Erica Naone that his algorithmically-driven trading firm made NO erroneous trades on the day of the flash crash. 

How did Phase Capital use algorithms and provide a safe trading harbor to their investors on the day of the flash crash? They use the same technology to analyze their market data in real-time that they use to trade in real-time: complex event processing, or CEP (intro here).

Corwin Yu, director of electronic trading for PhaseCapital, explained their innovation this way:

Typically, PhaseCapital processes 30,000 to 40,000 'ticks,' or pieces of market data, per second. During the flash crash, that number jumped to more than 289,000 ticks per second— much of it representing stocks swinging wildly. "A lot of scrubs kicked in and realized that this data didn't make sense," Yu says. For example, some large publicly held companies, such as Accenture, traded for less than $1 during the flash crash. PhaseCapital continued trading, but filtered data that appeared suspicious. Because it didn't act on that data, it was spared erroneous trades.

Yu concluded that speed is essential in the world of trading today, but that the trick is to be "fast and smart." Read about PhaseCapital's innovation in MIT Technology Review.

The solution for a safer market isn't heavy-handed regulation of "wild, reckless algorithms" in order to catch the "scoundrels" of Wall Street. Fast does good on Wall Street; but fast and dumb is dangerous. So we should tune out the fear mongers and tune in to the innovators that can help make Wall Street be fast - and smart.

Thomson Reuters Selects StreamBase CEP for Elektron High-Speed Cloud, Quant Platform, Apps, and FX

Today we're excited to report that Thomson Reuters has selected StreamBase as its complex event processing (CEP) platform for a broad set of initiatives. CEP "flips the database on its head" by providing the ability to monitor, analyze, and act on events happening in real-time, rather than looking in the rear-view-mirror of what happened yesterday, last week, or last month. CEP powers real-time, big data applications on Wall Street including high frequency trading, pre-trade risk management, surveillance, market data management, smart order routing, pricing, settlement, real-time P&L, monitoring - applications from front to back office on Wall Street.

Why did Thomson Reuters partner with StreamBase for CEP? Initially, they announced four reasons: 

StreamBase in Elektron Brings CEP to Thomson Reuters' High-Speed Cloud

The StreamBase platform lies at the heart of a new generation of real-time, automated applications on Wall Street. As the use of CEP continues to grow, our customers increasingly ask to deploy some of their applications in the cloud. They want their automated processing to happen as close to the exchange as possible for the lowest possible latency, and they want secure and reliable access to their market data.

Thomson Reuters announced that they will provide StreamBase CEP in Elektron, their open, global, ultra-high speed and resilient network for the financial industry. By adding StreamBase CEP to Elektron, Thomson Reuters is creating the industry’s only platform that includes comprehensive real-time and historical data, high capacity network connectivity, managed hosting services, and CEP for rapid application development and strategy deployment.

The First End-to-End Quant Platform for Back Testing, Algorithmic Development and Deployment

Although StreamBase is all about acting on what's happening now, the knowledge of what happened yesterday is essential for any automated system. History is essential to discover trends and patterns in event streams. History is essential to back test automated strategies. History is essential to gain perspective on real-time actions at execution time. Over 90% of applications that use StreamBase also use some form of historical event data management.

Thomson Reuters will integrate StreamBase with the Enterprise Platform for Velocity Analytics, which integrates all Thomson Reuters content, including tick history, reference data, corporate actions, condition code processing and cancellation and correction messages. It also brings together many disparate third party applications including data management solutions, EMS, OMS, price discovery, back testing, trade surveillance and low latency execution capabilities across the quantitative research and trading workflow.

By integrating Velocity Analytics with StreamBase CEP, Thomson Reuters has created the first end-to-end quant platform for back testing, algorithmic development and deployment. 

Thomson Reuters Applications "Powered By" StreamBase

Thomson Reuters will deliver a suite of applications that are “powered by” StreamBase CEP to address low-latency workflow. CEP inside will provide Thomson Reuters clients the benefit of time to market usually associated with turnkey solutions, while at the same time allowing full customizability of the underlying application. One application Thomson Reuters already provides is the real-time surveillance application from Redkite financial which is based on the StreamBase CEP platform. 

StreamBase and Thomson Reuters FX Solutions

The FX markets have rapidly emerged as an important market for the capabilities of CEP. Thomson Reuters announced that StreamBase CEP will complement the existing Thomson Reuters Dealing Aggregator FX trading solution, available exclusively to Thomson Reuters Dealing customers, by providing CEP capability in tandem with the low latency algo trading platform. 

Why did Thomson Reuters choose StreamBase? Mike Powell, Global Head of Enterprise Real-Time Solutions for Thomson Reuters, said it this way: “The financial market landscape continues to evolve rapidly, meaning that nimbleness and time to market have become essential for market participants. Elektron empowers our customers to innovate and achieve high performance through a seamless infrastructure that brings together global market participants, delivers world-class content and creates a suite of new added-value services. The combination of StreamBase’s market leading CEP engine together with Elektron will provide a platform for us to deliver flexible and sophisticated solutions to our customers faster and more efficiently.”

For StreamBase, we’re excited about the vast resources and capabilities of Thomson Reuters can provide our customers. We’re honored to be selected as Thomson Reuters’ CEP platform of choice - their selection validates our focused approach to CEP, the ever increasing importance of the cloud and real-time big data, the disruption of simplicity, and the continued need for low latency and scalability on Wall Street. 

 

Can Innovative Surveillance Be a New Source of Competitive Edge?

In sports, great defense can be a catalyst for great offense. Can the same idea hold true for surveillance - the defense - in the capital markets? This was a question raised by today's panel of surveillance experts at today's StreamBase webinar (watch the replay). 

First, TABB Group's Miranda Mizan assessed the current surveillance market. She described how surveillance is morphing from a defensive obligation to an offensive differentiator, because the markets are changing so dramatically. Mizan argues that the quantity of data, speed of trading, and complexity of instruments have made traditional surveillance techniques insufficient. And since regulation (e.g., MIFID II) is changing so quickly, she argues that firms must re-think their surveillance strategies and technology choices.

Justin Amos joined Miranda. Justin is a former global markets sales and trading executive at JPMorgan, and the MD of Redkite Financial Systems, a UK-based firm whose system called Redeye has been quickly becoming the standard for modern, real-time surveillance systems. Justin described the requirements of today's surveillance systems by describing the motivation for Redeye, which he likened to "anti-virus software for trading."

Together, Justin and Miranda touched on 12 "next generation" surveillance requirements, including:

1. Real-time analytics. Surveillance systems must analyze market data at the speed of automated trading.
2. Historical "big data" analytics. On-line access to trade-by-trade history for investigation, trending, and discovery.
3. Data fusion. Mash-ups of increasingly complex data, including unstructured data.
4. Inclusion of new data sources. The ability to integrate live data, static data, and social media data.
5. Multi-asset class support. Modern surveillance must work across asset classes.
6. Pre-built surveillance algorithms. Standard algorithms like front running, spiking, tailgating, and ramping.
7. Adaptive algorithm modification. The ability to adapt algorithms on the fly to reduce false positives.
8. Alert algorithm extensibility. The ability to quickly add new algorithms to the system as market dynamics evolve
9. Customizable surveillance algorithms. The ability for a firm to tweak an algorithm for their own models.
10. Flexible workflow management and case management. Control over the process of alert analysis, resolution, and tracking.
11. Accessibility for business analyst. With regulation changing so quickly, more firms want their own control over surveillance behavior.
12. Cloud-based or on-premise deployment. The cloud allows more rapid rollout, and may ease global scalability challenges

The panel argued that existing surveillance systems can't handle these requirements. Taken together, the challenge makes what was once traditional "defense," or a compliance function, very new and challenging - a source of differentiation.

Miranda went deeper, with examples of how brokers and the buy side are using modernized surveillance systems: agency brokers are beginning to use surveillance tools to monitor their trading algorithms and optimize their client service and order routing for best execution; the buy side are using surveillance to gain greater control to know that their trading is safe, its orders get to the right places, and that dark pools are working correctly in relation to the rest of the market; sell side institutions are using surveillance tools to institute process and control across the organization, globally, and integrated with automated systems.

Richard Tibbetts, CTO of StreamBase, gave a brief introduction of StreamBase CEP, and why complex event processing is increasingly being used to power surveillance applications; indeed, CEP provides the engine to address many of the 12 modern surveillance requirements set out by the panel.

Simon Townsend, CIO of Redkite, gave a "whistle stop tour" around Redkites' application. He hit on most of the 12 requirements above, and even showed didn't just claim that their algorithms are extensible and dynamic (requirements 6, 8 and 9) - he showed it in StreamBase Studio, and showed how business analysts can extend the Redeye surveillance algorithms.

One of the most exciting times is when a market is transformed by new ideas. In the capital markets, perhaps that innovation is upon us in surveillance, where tools that meet modern trading requirements can turn a defensive obligation into a tool for innovation.

Twitter and Trading: A Ripple in the Big Data Wave

In 2009, StreamBase began talking publicly about using CEP with Twitter. The topic sparked a flurry of mainstream media coverage about Complex Event Processing (CEP) on Wall Street on outlets such as CNBC, the Wall Street Journal, and the Financial Times. Even Popular Mechanics named CEP as one of 10 Tech Concepts to watch, along with Fracking and Flywheel Energy Storage! (watch this short video for an introduction to CEP)

Last week, USA Today published an article about CEP and trading on the front page of its Money section: Wall Street traders mine tweets to gain competitive edge. The report covered developments such as: 

• The opening of Derwent Capital, a firm that specializes in news-driven trading strategies, including Twitter
• A report from Adam Honore from AITE group whose two recent reports (on trading strategies and on unstructured data) revealed that interest in unstructured data has jumped from 2% in 2008 to 35% in 2011
• Rich Brown, who talked about how the Thomson Reuters news sentiment engine is more effective than the "quirky and abbreviated language of Twitter"
• Johan Bollen, a professor of informatics at Indiana University, whose research claims an 87% accuracy rate in using Twitter mood measurements to predict Dow stock prices three to four days later
• MarketPsych, a firm that used to run sentiment-based strategies, switched from being a hedge fund to a software provider

The article shows that Twitter continues to be a controversial topic: some tweets are clearly important - for example, the CME Group tweets news about the commodities market and has over 780,000 followers. On the other hand, computer-measured sentiment is just that - one more barometer of what the masses are thinking. Good investment decisions are powered by experience and knowledge, and sentiment is one input of many.

The noise over use of Twitter on Wall Street drowns out the real story, which is why I was quoted in the article that Twitter is "a ripple, not a wave" on Wall Street (and Main Street as well).

The big wave is big data, and Twitter generates just a few buckets of sand in world of big data today. 

In "New Rules for Big Data," the Economist reported that Google processes a petabyte of data an hour. Here's a way to put Twitter in context*: think of a bucket of sand. If every grain of sand in the bucket was 1 byte of data, then:

• The entire work of Shakespeare fills just one bucket of sand (about 5MB)
• A fast financial market data feed (OPRA) fills a beach of sand in 24 hours (about 5TB) 
• Google processes all the sand in the world every week (about 100PB)
• We generate 60% more sand every year

In these terms, all of Twitter generates only a sand castle of quality data a day. While some of that data is very valuable, such as tweets from CME Group (@CMEGroup), most of it simply tells us about mass sentiment.

On the other hand, the debate about Twitter does reveal the opportunity and challenge of big data. The Economist used a quote from Oscar Wilde in 1894 to put that challenge in context - Wilde said: “It is a very sad thing that nowadays there is so little useless information."

So the big ruckus about Twitter on Wall Street reveals the real big story - today's big data challenge.

* Thanks to Rob Doherty at Cloudswitch for sharing Chad Sakac's (EMC) buckets of sand concept, the Economist New Rules for Big Data for sizing data from Google, and @RJMetrics on big data (TED Philly talk) for conceptual description of big data.