Optiver Australia Pty Ltd v Tibra Trading Pty Ltd

REASONS FOR JUDGMENT 1 This is an application by the applicant ("Optiver") for preliminary discovery pursuant to O 15A r 6 of the Federal Court Rules 1979 (Cth) ("the Rules"). The application and the subsequent notice to produce seek documents from the respondents ("Tibra") relating to computer programs written by employees or former employees of Optiver during their employment with Optiver, and other related information.

THE PROCEEDINGS 2 Optiver is a proprietary arbitrage business. It buys or sells securities in different markets in order to capture a profit on the price differences between those markets. Its modus operandi is to identify securities on Australian and foreign markets which appear to be mispriced, and then buy or sell those securities for a profit. This process has the effect of reducing anomalies in those markets by trading the relevant securities back into equilibrium. Optiver conducts this business for its own profit, not for the profit of clients, and has done so in Australia since 1997. 3 The respondents also conduct a proprietary arbitrage business. The first to fourth respondents were each incorporated in 2006 or 2007, and their directors and employees are drawn variously from the fifth to ninth respondents, all of whom were previously employees of Optiver. 4 Arbitrage firms rely on speed. Due to the speed with which trades take place, any hesitation or delay by the trader or the trader's computer software can result in lost trades. Arbitrage computer software is designed to identify price anomalies, to place trades which will yield a profit, and to do so at a speed which will beat other traders to those trades. 5 Optiver's computer software has gone through several stages of development. When it began trading in 1997, Optiver used an "off-the-shelf" software product known as "Orc". This program was capable of responding to market data in about 80 milliseconds. Over time, Orc was superseded by other software. In response to improving "off-the-shelf" software, Optiver sought to program its software so that it would be faster than any other program. 6 In early 2004, employees of Optiver wrote a source code for a piece of software which would operate in conjunction with Orc. It was known within Optiver as "CATS". This program dramatically reduced Orc's response time to between 0.5 and 1.5 milliseconds. 7 Despite the success of CATS, Optiver continued to produce faster software. In late 2004, several of Optiver's employees developed an enhancement of CATS. The product was called "F1", and was regarded by Optiver as crucial to the success of its arbitrage business. F1 is extremely fast, utilises complex algorithms which allow it to "think like a trader", and contains in-built mechanisms to minimise losses resulting from market or human errors. 8 In late 2005 and early 2006, Optiver's staff structure changed. Optiver terminated the employment of the fifth respondent, Mr Dinesh Bhandari, in November 2005. By June 2006, six further employees of Optiver, including each of the sixth to ninth respondents, tendered their notices of resignation. Optiver's evidence as to the circumstances surrounding these resignations - adduced in two affidavits of Mr Robert Keldoulis, Optiver's Managing Director, and one affidavit of Mr Charles Shale, a Software Developer at Optiver - details the way in which Mr Keldoulis formed suspicions that the fifth to ninth respondents were leaving to establish a company in competition with Optiver. 9 Soon after their respective resignations, each of the fifth, sixth, seventh and ninth respondents either incorporated or were appointed as director of at least one of the first to fourth respondents. Mr Andrew King, the eighth respondent, was employed by Tibra. 10 In July 2006, the first respondent was registered with the Australian Stock Exchange as a "market maker", which allows it to conduct an arbitrage business in the same way that Optiver does. Optiver's evidence was that Mr Keldoulis became suspicious at the speed with which Tibra had moved from a newly incorporated company to a successful competitor. The evidence indicated that Mr Keldoulis discovered emails sent from some of the fifth to ninth respondents' email addresses at Optiver to their personal email addresses. One of these emails, sent by Mr King, itemised ways in which Optiver's F1 system could be improved. The other, sent by Mr Kinsey Cotton, the ninth respondent, contained key codes for Orc, for which Optiver had acquired licences and on which Optiver had built its F1 automated trading system. The information contained in these emails, says Optiver, is not only information crucial to a successful arbitrage business, but also information which the fifth to ninth respondents, who were traders and not software developers, could not have easily recreated. 11 At times in cross-examination Mr Keldoulis appeared evasive, and the correctness of some of the details of his affidavits was undermined, and consequently I consider that his evidence reflects an eagerness to advance his claims and an unwillingness to reveal Optiver's position in an untimely manner. However, despite this, I accept the substance of his evidence. On the material presently before me, I do not consider that that substance has been materially shaken. 12 Optiver also adduced evidence from Mr Shale concerning the performance of Tibra as an arbitrage business and the speed with which a software developer could develop a program competitive with F1. 13 In relation to the performance of Tibra, Mr Shale's evidence was that Tibra began beating Optiver to an increasing number of trades in late 2006. Although the evidence deposed to in his affidavit addresses only the period between September 2006 and December 2006, a graph annexed to his affidavit asserted that Tibra's rate of success over Optiver continued to increase into early 2007. Based on his experience as a software developer, Mr Shale concluded that Tibra would need to have deployed an automated trading system at least as fast, if not faster, than Optiver's F1. 14 Furthermore, Mr Shale deposed that, for a skilled software developer to generate a source code with functionalities equivalent to F1, the developer would need, even taking into account several assumptions in favour of the developer, at least 110 weeks of development time. Having concluded that Tibra began using a software program equivalent to F1 in November 2006, Mr Shale concluded that it was highly unlikely that Tibra could have developed such software by that time. Although his evidence was challenged in cross-examination because it did not clearly set out the time necessary to develop CATS as distinct from F1, I consider that the substance of Mr Shale's evidence was not significantly shaken. 15 Tibra did not lead any evidence in response to that adduced by Optiver. It tendered a letter sent by Tibra's solicitors to Optiver's solicitors, responding to a request by Optiver that Tibra voluntarily surrender certain relevant documents, source code and computer systems. Tibra's response addressed each of the substantive arguments put against it in the preliminary discovery proceedings. 16 In the absence of the calling of a witness to explain the information contained in Tibra's letter, it is difficult to give the document great weight or probative value. It suffices to observe that Tibra, through their solicitors in this letter, formally rejected Optiver's suspicions of breach of confidence and copyright and refused to produce the documents, source code and computer systems listed in the application.