`
`————————————————
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`————————————————
`
`GAIN CAPITAL HOLDINGS, INC.,
`Petitioner,
`
`v.
`
`OANDA CORPORATION,
`Patent Owner.
`
`————————————————
`
`Case No. CBM2020-00021, Patent No. 8,392,311
`
`————————————————
`
`DECLARATION OF DR. MICHAEL STUMM
`
`OANDA – EXHIBIT 2005
`
`
`
`I, Dr. Michael Stumm, hereby declare:
`
`
`1.
`
`I am a professor in the University of Toronto’s Department of Electrical
`
`and Computer Engineering. I have published over 100 papers in top-tier conference
`
`proceedings and scientific journals.
`
`2. My research interests lie in the general area of what in the computer
`
`science community is referred to as computer systems, particularly multiprocessor
`
`and distributed systems. One focus of my research and study is the design and
`
`engineering of distributed systems. I, along with the students in my research group
`
`at the university of Toronto, have built from scratch both computer operating
`
`systems, designed as distributed systems, as well as the hardware they run on,
`
`including 16-processor and 64-processor shared memory systems.
`
`3.
`
`Generally, a distributed system is a type of computer system which
`
`amalgamates multiple computer systems together by locating various components
`
`of the system that are on different physical computers and which communicate with
`
`each other, for example by passing messages to each other. These different
`
`components interact with each other to achieve a common goal.
`
`4.
`
`I am the inventor or co-inventor on at least eleven U.S. patents related
`
`to market and currency trading and telecommunications networks including U.S.
`
`Patent Nos. 7,146,336 (“’336 Patent”) and 8,392,311 (“’311 Patent”).
`
`
`
`1
`
`
`
`5.
`
` In 1996, I, along with Dr. Richard Olsen, launched OANDA, a
`
`company that provided the world’s largest and most accurate database of currency
`
`prices at that time. OANDA soon became the gold standard for currency prices and
`
`interbank exchange rates online—relied upon by major corporations, national central
`
`banks, the United States Internal Revenue Service, auditing firms, and individual
`
`traders alike.
`
`6.
`
`Although OANDA had made accurate exchange rates more available
`
`to the public, there remained a lack of viable platforms for individual retail traders
`
`to trade currency pairs (also known as foreign exchange, “forex,” or “FX”). At that
`
`time, to trade currency pairs, retail traders had limited options, each with their own
`
`drawbacks. First, retail traders could try to go through banks and currency dealers,
`
`but these charged consumers large spreads when trading currency, i.e., when the bid
`
`and ask prices are significantly far apart. And while some online trading platforms
`
`existed at that time, they suffered from a number of technical deficiencies. For
`
`example, due to the inefficient construction, traders could not see the prices of
`
`different currency pairs change dynamically—a user had to refresh their browser
`
`window to get new prices. In an attempt to address this shortcoming, some platforms
`
`constructed their web pages to automatically refresh once a minute, which would
`
`then show new prices. However, a browser refresh was (and still is) highly
`
`disruptive to the user. For example, if the user had scrolled down the page, they
`
`
`
`2
`
`
`
`would lose their place; and information the user had partially entered into forms
`
`(such as orders) could be lost. Additionally, automatic page refreshes more
`
`frequently than once a minute were prohibitively resource intensive for the trading
`
`platforms because, typically on a browser page refresh, the entire page had to be
`
`resent from the server to the browser, even though often the only thing that actually
`
`changed was the prices of the currency pairs. Additionally, retail systems before
`
`OANDA’s could not support continuous monitoring of pending order positions
`
`(such as stop loss or take profit orders), with these often being updated only
`
`infrequently or overnight.
`
`7.
`
`To make online currency trading systems more useful for retail
`
`customers, OANDA invented systems and methods for online currency trading that
`
`overcame these and other deficiencies of then-existing online currency trading
`
`technologies.
`
`8.
`
`In 2000, I helped work on designing and building an online automated
`
`trading platform, through which OANDA could offer retail investors more favorable
`
`rates that banks used to trade currency among themselves. One way that we
`
`overcame the technical deficiencies inherent in prior art online implementations was
`
`through the use of a relatively new technology, Java. By creating and using carefully
`
`constructed client-server systems using Java applets (where some code executes on
`
`the user’s computer (in the Web browser), and some code executes on the trading
`
`
`
`3
`
`
`
`platform server or servers) we could achieve second-by-second updates without
`
`costly whole page refreshes.
`
`9.
`
`Notably, an online currency trading platform is neither merely a
`
`“computer” nor is it merely “software.” Rather, an online currency trading platform
`
`is a specialized type of distributed system, comprising (i) server hardware (including
`
`physical computer servers and databases in a datacenter), (ii) server software
`
`(including various programs that configure the computer servers and databases to do
`
`their jobs, cooperate, and communicate), (iii) networking equipment, (iv) client
`
`hardware (including customers’ desktop computers), and (iv) client software
`
`(including the Java code that runs on the customers’ desktop computers).
`
`10. The design and assembly of the various components into a serviceable
`
`distributed system improved the functioning of the individual computers, both the
`
`server and consumer side machines, themselves. For example, a prior art trading
`
`platform server, like the ones discussed above, that served static pages (i.e., those
`
`requiring a refresh to see new prices) could only serve a small fraction of the number
`
`of customers that a dynamic trading platform server could service. Systems based
`
`on static pages relied on the server hardware to do all of the work necessary to update
`
`the client system with a new price. However, combining the client hardware with
`
`the server hardware into a distributed system reduced the amount of work that the
`
`
`
`4
`
`
`
`server computer had to do to achieve this same goal, thereby allowing the distributed
`
`system to service a larger number of customers without added costs.
`
`11. To illustrate just one of the technological solutions to this problem that
`
`OANDA invented, we were able to improve on several of the deficiencies of pre-
`
`existing systems by using the previously mentioned Java Applets. Because we were
`
`able to build our client-side software using Java Applets to establish a persistent
`
`connection with the trading system server(s) and only send updated pricing
`
`information rather than entire pages, we eliminated the latency and inefficiency
`
`associated with page refreshes. Additionally, our unique and non-standard use of
`
`Java Applets to create a client-side trading system also avoided the downsides of
`
`traditionally installed client-side applications, which were difficult for users to
`
`install, created lots of client-side technical support issues with firewalls, and only
`
`worked on specific operating systems such as Microsoft Windows. In contrast,
`
`OANDA’s browser-based Java Applets ran in practically any browser on any
`
`operating system, did not need network or firewall configuration on the user’s side,
`
`and did not require end-user installation or upgrading.
`
`12. OANDA’s first fully automated online currency trading platform,
`
`fxTrade, was groundbreaking in its use of a distributed system like the ones
`
`described above. As a result, among other features, fxTrade monitored market
`
`exchange rates, offered immediate price quotes (with a much smaller spread than
`
`
`
`5
`
`
`
`offered by banks), executed trades instantaneously, and prevented clients from
`
`risking too much money through automatic stop-loss orders. It also allowed
`
`customers to trade with deposits as small as one dollar, while charging interest on
`
`leveraged trades on a second-by-second basis.
`
`13. These inventions and their implementation in the fxTrade platform
`
`allowed users (our clients) to see price fluctuations without clicking “refresh” or
`
`having to take any action on their own. In fact, the fxTrade platform provided prices
`
`both in text listings as well as visual graphs—an entirely new development in the
`
`market at the time we filed our patent applications.
`
`14. One
`
`important
`
`feature of OANDA’s platform, which was
`
`unprecedented and not available on competitors’ platforms, was that OANDA’s
`
`system was sufficiently fast and efficient to provide real time prices that traders
`
`could actually trade on. On competitors’ systems, when a customer requested to
`
`trade on certain currencies, the system would take up to a few seconds to process the
`
`trade and to confirm that the price was favorable to the bank. If favorable to the
`
`bank, the trade would be accepted (at the customer’s loss). If not favorable, the trade
`
`would be rejected and a new quote generated, which would force the customer to
`
`request the trade a second time.
`
`15. At
`
`the
`
`time, all of OANDA’s competitors—including banks,
`
`companies, and trading and investment houses—were unable to achieve second-by-
`
`
`
`6
`
`
`
`second price updates like those on OANDA’s platform. I recall a meeting with
`
`executives from Morgan Stanley who expressed their admiration of OANDA’s
`
`groundbreaking developments. In fact, the Morgan Stanley executives noted that
`
`they had hired over 50 programmers to try and reverse engineer the OANDA
`
`platform; but they were unable to do so and gave up.
`
`16. One surprising thing we learned in designing and building the OANDA
`
`trading platform was the necessity of integration. In other words, a functional,
`
`efficient platform required a single distributed system that cooperated as a single
`
`unit, rather than multiple systems that were simply connected to each other.
`
`17. At the time OANDA developed its fxTrade platform, the common
`
`belief and practice in the industry was that the best way to create these sorts of
`
`currency trading systems was to select “best of breed” components and combine
`
`them, i.e., select a top-of-the-line pricing engine and combine it with a good graphing
`
`engine, etc. Contrary to this conventional wisdom, my team discovered that this
`
`practice resulted in more expensive, slower systems that could not meet customer
`
`demands in the open market. The problem with this approach was that the individual
`
`components were not designed to cooperate, did not inherently know how to talk to
`
`each other, did not have compatible interfaces, and simply did not integrate well. To
`
`make the separate components fit together, the system designer had to create
`
`numerous “glue” components to join the system pieces together. As one example,
`
`
`
`7
`
`
`
`the designer would have to translate protocols or language of one system to the
`
`protocols or language of another system. These “glued” components and steps
`
`resulted in a system that was slow and inefficient.
`
`18. Of high importance to OANDA’s retail trading platform business was
`
`maintaining highly accurate real-time prices, to prevent customer arbitrage.
`
`OANDA’s retail business model at that time did not charge customers commissions.
`
`Instead, OANDA made its money from the spread between the “buy” and “sell”
`
`prices of each asset. Spreads were typically very small and measured in units of
`
`1/100 of a cent. Thus, it was important for OANDA’s prices, displayed and updated
`
`in sub-second frequencies on its customers’ computers, to overlap the market prices
`
`in real time, or OANDA’s customers would arbitrage them. In other words, if
`
`OANDA’s “buy” price was higher than another company’s “sell” price, a customer
`
`would simply buy from OANDA’s competitor and sell to OANDA. Or, if
`
`OANDA’s prices updated slower than its competitors’ prices, customers could
`
`watch the competitors’ prices to see the “future,” and then use that knowledge to
`
`quickly buy and sell on OANDA’s platform, at OANDA’s expense. The techniques
`
`and methods developed at OANDA to accurately calculate and predict the prices to
`
`be displayed to consumers was highly complex and a huge part of OANDA’s ability
`
`to offer an online real-time retail trading platform.
`
`
`
`8
`
`
`
`19.
`
`In the retail FX trading business, slower is often not just an
`
`annoyance—it made the system unworkable. As discussed, if a system is too slow,
`
`then customers can take advantage of the delays to arbitrage the trading platform.
`
`Because of the speed and accuracy of OANDA’s real-time prices, OANDA was also
`
`one of the first trading platforms to offer its customers slippage forgiveness without
`
`concern over affecting OANDA’s trading profits. Slippage is the change in the
`
`currency exchange rate from when a trade is first requested to when it is completed.
`
`OANDA’s competitors would use slippage to increase profit when slippage was in
`
`its favor or refuse a trade when slippage was not in its favor.
`
`20. Because of OANDA’s advances in the technology, OANDA’s trading
`
`platforms and systems were able to accept what was technologically and practically
`
`speaking a new type of order—an order for immediate execution at a specified price.
`
`The order price was set by a user interface field which would be pre-populated with
`
`the current price as continuously received in real time from OANDA’s servers.
`
`Different from a traditional “market order,” OANDA’s real-time order specified a
`
`price which was, by default, set to be the price shown on the user interface at the
`
`time the order was placed, and an order sent with that price was guaranteed to
`
`successfully execute at the stated price if the order arrived at the servers within a
`
`reasonable time (e.g., a few seconds). OANDA’s real-time order is also different
`
`from a traditional “limit order” because it is executed immediately on receipt, not
`
`
`
`9
`
`
`
`held for later. Practically speaking, from the perspective of the trader and the system
`
`operator, OANDA’s real time order was a fundamentally different trade because it
`
`allowed the trader to see a price immediately execute an order to trade at that specific
`
`price.
`
`21. Because of the efficiency of OANDA’s distributed hardware/software
`
`systems, OANDA became a market maker because it opened up the market to higher
`
`frequency trading, in smaller quantities (down to $1 trades). Also, because of the
`
`relative efficiency of our distributed hardware/software systems, which made the
`
`component computers function better, OANDA also expanded the market from 4
`
`digits to 5 digits of precision in the currency prices, or what was called at the time a
`
`“pipette.” No banks or other competitors could offer this at the time.
`
`22. Another innovation that OANDA’s systems enabled was real-time
`
`interest payments. Generally, an investor holding a currency position wants to be
`
`paid interest on that currency. With other platforms, interest would be calculated
`
`and paid based on how much currency the investor held at 5:00 p.m. (the end of the
`
`trading day). But, because of OANDA’s efficient real-time pricing and interest rate
`
`engines inventions, OANDA’s accounts could calculate and pay interest on a
`
`second-by-second basis.
`
`23. Additionally, in the then-existing currency trading market, currency
`
`traders and/or trading platform operators determined value-at-risk—an important
`
`
`
`10
`
`
`
`number for maintaining margins and avoiding forced sales—using pre-existing prior
`
`art methods such as “daily data” calculations, or what were known as RiskMetrics
`
`methods. These methods inherently resulted in problematic high stochastic error as
`
`they relied on homogenous time series data despite the fact that financial markets
`
`produced inhomogeneous data (irregularly spaced in time).
`
`24. The improved methods OANDA developed for determining value-at-
`
`risk in online trading platform systems were completely novel. Typically, in FX
`
`trading, customers are highly leveraged, meaning that for every dollar a client
`
`deposits with a trading company, they may be trading $20, $50, or even $100 worth
`
`of assets. Thus, if a client traded into a losing position, the client’s account balance
`
`could quickly go negative. Having to ask customers for more money after bad trades
`
`was a quick way to lose customers, so OANDA developed and implemented stop-
`
`loss limits to protect against customers falling into in the negative. Calculating each
`
`customer’s value-at-risk in a manner that allowed OANDA to apply the stop loss
`
`limits was a highly intensive and innovated process—each customer’s account value
`
`had to be constantly evaluated and re-calculated second-by-second, using models
`
`and projections of the current prices and interest rates, to protect the profitability of
`
`our retail trading businesses.
`
`25. The way that OANDA’s trading platform systems performed hedging
`
`was also new. Whenever a client purchased foreign assets (for example, 100€),
`
`
`
`11
`
`
`
`OANDA would then be “short” that asset. Thus, to hedge against the risk of clients’
`
`purchases of foreign assets, OANDA would go to the currency markets and purchase
`
`that 100€ so that it had a balanced position. OANDA’s inventions allowed it to do
`
`this balancing in an efficient, feasible manner. To do this, we recognized that it
`
`would have been inefficient to constantly make many small purchases in the
`
`currency markets and also that many of OANDA’s individual client purchases were
`
`offsetting each other. In other words, if one client bought 100€ but another client
`
`sold 200€, OANDA’s net exposure is -100€. By using a server component of the
`
`distributed trading platform system known as a hedging engine, OANDA efficiently
`
`consolidated many customer transactions together to determine net exposure and use
`
`those consolidated amounts to hedge risks more efficiently from its customers.
`
`26. As trading moved online and became more high frequency—due to
`
`OANDA’s inventions—it was important to use more accurate, real-time methods
`
`that could take advantage of the inhomogeneous data. The need to process
`
`inhomogeneous data arose because now traders could trade whenever they wished,
`
`as opposed to at a fixed time governed solely by the dealers. For trading to become
`
`high frequency, computerized, and automated, data streams from different sources
`
`had to be integrated, and bad data had to be filtered out of the system. Humans alone
`
`could not process the data fast enough or accurately enough to use it in a real-time,
`
`accurate manner. In fact, one of the reasons computerized filtering is necessary is
`
`
`
`12
`
`
`
`to filter out erroneous data that is inputted by humans. With the availability of high-
`
`frequency trading data and an online trading platform, new short-term trading
`
`strategies became available to both institutional and retail traders.
`
`27. Along with this, the increasing availability of cheap computing power
`
`in the early 2000’s created a demand for trade recommendations based on real-time
`
`price data and position information. Existing prior art methods to get this
`
`information (e.g., calling a broker or keeping track of assets and options by hand)
`
`were too slow, inaccurate, and impossible to use in a real-time online trading space.
`
`OANDA’s inventions of asset trading using purpose-built trading recommendation
`
`calculators and predictive methods enabled traders to benefit from these real-time
`
`predictive engines, and platform providers to benefit by offering them to their
`
`customers. Some of the inventions disclosed in these patents were embodied in
`
`OANDA’s pioneering currency trading platform, fxTrade, which launched in 2001.
`
`28. At bottom, OANDA’s inventions and its fxTrade platform embodying
`
`those inventions resolved many issues prevalent in the foreign currency trading
`
`market that acted as roadblocks to most retail customers looking to trade foreign
`
`currencies online. OANDA’s developments provided for an efficient, real time
`
`trading platform that allowed retail customers to act on real-time pricing and trade
`
`in much smaller amounts than what was possible on other platforms. The import of
`
`OANDA’s development was evidenced by its success—we were a tiny company
`
`
`
`13
`
`
`
`that, with our inventions, was able to execute more trades per day than large banks
`
`like Deutsche Bank or UBS. Not only did OANDA execute more trades than its
`
`competitors, but its innovations also changed the nature of online FX trading itself.
`
`29. All of the statements made in this declaration of my own knowledge
`
`are true and all statements made on information and belief are believed to be true.
`
`These statements were made with knowledge that willful false statements and the
`
`like so made are punishable by fine or imprisonment, or both, under section 1001 of
`
`Title 18 of the United States Code.
`
`Executed on this December 15, 2020, at Toronto, Canada,
`
`Dr. Michael Stumm
`
`14
`
`