Trading titans’ clash swords in discovery clash

David McGrath follows the trail of a single email which is at the heart of a two year court battle between electronic trading giant Optiver and Aussie rising star Tibra.

This year Dinesh ''Danny'' Bhandari debuted on the BRW Young Rich List with a personal fortune conservatively estimated at $A40 million. Last year he was named young entrepreneur of the year by Ernst & Young.

Danny heads up a group of companies called Tibra which has enjoyed extraordinary success in the space of a few short years, turning a profit of $A14 million in its first year, $A57 million in its second and $A77.5 million last year. Not bad at all for a startup unfortunate enough to go into business just before a global financial crisis.

The secret to his success? Smart software – of the ultra fast, world class market trading kind. The kind of software used all over the world to rake in millions of dollars in profits every year from so-called market “imbalances” and “mis-pricing”.

The kind of software that, on some estimates, accounted for as many as 80% of US equity market trades in 2008 and which has spawned a “technological arms race” for market supremacy.
In industry parlance, it is known as algorithm trading, or simply “algo trading”

‘Algo trading (aka ‘black box’ or ‘robo’ trading) uses sophisticated software to automate market trading. As it is automated, it can analyse, test, and trade multiple markets at speeds that mere mortals simply cannot compete with. This has led to new trading styles such as “high frequency”, “low latency” and “iceberging”.

These techniques are widely used in a variety of financial markets including stock, bond, foreign exchange, options and futures. They are used by a range of players from conservative investors such as pension and mutual funds, through to retail and investment banks and at the more speculative end hedge funds, arbitrage traders and ‘market makers’.

It is causing headaches for industry regulators trying to keep an open and level playing field for all market participants.

There is no doubt that Danny and his team at Tibra are very good at this. Their software was beating competitors to crucial trades just months after the firm was setup.

Danny’s former employer, Optiver smelt a rat, and for the past two years has been pursuing Tibra through the courts, alleging theft of its intellectual property.

To determine whether Optiver's claims are correct, it says it is necessary to examine the "source code" of Tibra's trading algorithms and compare the programming to its own.

Tibra is however strenuously resisting these efforts, and is understandably less than interested in handing over the secrets to its phenomenal success to a key competitor.

There have so far been seven judgments and appeals all the way to the High Court, and the case is by no means over. In fact, Optiver freely admits that it can’t decide whether it has a case against Tibra, until it sees its source code.

So what have they been arguing about all this time? Preliminary discovery of Tibra’s source code. Once again, discovery proves just how central it can be to court action.

Exciting Optiver’s suspicions were some emails from its employees, sent to their personal email addresses shortly before resigning to join Tibra.

One of those emails has proved to be crucial - this email, sent by one of Tibra’s director’s, Mr King, to his personal account on the date of his resignation, itemised ways in which Optiver’s F1 system could be improved!

So how did it get to this?

The Master’s Apprentice Awakens?

Optiver has been in the “algo trading” business since 1997. Optiver uses its software for arbitrage trading.

Arbitrage trading takes advantage of a price difference between similar or related securities on different markets. These price differences are tiny and are quickly closed by the arbitrage transaction itself, so players need to be lightning fast. To the victor go the spoils - a small but relatively certain and risk-free profit margin. With the very large volumes transacted, these small margins turn a very tidy profit` for the trader.

In 1997, Optiver used an off the shelf product called ‘Orc’. Orc could respond to market data in about 80 milliseconds. As competitor speeds improved, Optiver wrote further code inhouse to improve response times.

In early 2004, Optiver deployed its own bolt on to “Orc” called “CATS”. This reduced response times to between 1- 1.5 milliseconds. Later that year, “CATS” was superseded by the aptly named “F1” (or ‘Formula 1’). F1 was extremely fast. In addition, it was able to mimic “trader thinking” and even included safeguards to minimise losses from errors. F1 was regarded by Optiver as crucial to its success. By late 2005 there was trouble brewing in the Optiver camp. Optiver became suspicious that some of its employees were planning to leave and setup a rival company. In November, 2005 Optiver terminated Danny Bhandari’s employment. Then followed a walk out by another six employees over the next six months. Most became directors or employees of the new company, Tibra, which Danny had set up with an initial investment of $2 million.

Things moved quickly for Tibra. By July, 2006 it was registered with the stock exchange as a ‘market maker’ enabling it to compete with Optiver. By the end of 2006, the newly formed company was beating Optiver at its own game – by being faster to the punch on the crucial arbitrage trades. According to expert evidence tendered at one hearing, the rate of success over Optiver continued to climb in 2007.

How was Tibra able to beat its old rival so quickly? After all, development of CATS and F1 by Optiver had taken years. Optiver engaged investigators to look into it.

Email evidence

They found emails that had been sent by Tibra staff when they were employed by Optiver from their work email accounts to personal email addresses. One email listed ways in which ‘F1’ could be improved. Another contained “key codes” for Orc, the original OTS software around which Optiver’s algo trading system had been built. Optiver alleged that this information was not only crucial to running an arbitrage business but that it could not have been recreated by the former employees who were traders not software developers.

Optiver also called evidence from an expert software developer. His evidence was that, in order to beat Optiver to these trades, Tibra needed to have an ‘automated trading system’ as least as fast as “F1”. Even after making a number of assumptions in favour of Tibra, he estimated that building such a system would require at least 110 weeks programming time. Given that Tibra was beating Optiver to these trades in November 2006 his evidence was that Tibra simply had not had enough time to develop this software.

In response, Tibra says that there are significant issues with the evidence presented by Optiver and that the computer expert’s evidence was far from conclusive. It disputed that the employment terminations were suspicious and maintains that its own success in the arbitrage business is not related to any misappropriation of Optiver’s confidential or copyright information.

The “Source Code” Judgment

These proceedings were strenuously contested from the beginning. Early on, Justice Tamberlin had ruled on two skirmishes where Tibra had tried to punch holes in Optiver’s case.
In the first, Tibra wanted Optiver to produce a number of categories of documents for the preliminary discovery hearing. These documents were to be used by Tibra to cross examine Optiver’s witnesses. Whilst the judge ordered Optiver to produce a small proportion of these, the vast majority he found were not relevant, even though they could be relevant to a final hearing in the matter. On the second, Tibra took various technical objections to Optiver’s evidence. In particular, it said that any references to Optiver’s source code were inadmissible unless it could prove copyright, which could only be done to producing the code.

The judge however ruled that on the preliminary discovery application it was appropriate to relax the normal rules of evidence. This spared Optiver the need to produce its source code.
In his 21 December, 2007 decision on preliminary discovery Justice Tamberlin found that Optiver already had sufficient information to decide whether or not to bring proceedings. He had in fact previously ordered pre-action discovery against Tibra. This time however Tibra was literally being asked to hand over the crown jewels i.e. all of its source code for its “algo” trading system, for inspection. His honour noted that this information was extremely sensitive and disclosure of it had the potential to substantially prejudice Tibra’s commercial interests by undermining its competitive advantage. While this could be mitigated by a confidentiality regime this would add cost, delay and complexity which was he felt was not warranted at that stage.

In addition, as a result of Optiver’s delay in bringing the proceedings, the source code now included Tibra’s improvements subsequent to the issues in the matter. Finally, he noted that as preliminary discovery was “one way” its effect in this matter would be to create an “inappropriate balance” between the parties during any subsequent litigation.

Justice Tamberlin stressed several times during his judgment that the purpose of preliminary discovery was not to strengthen a prospective plaintiff’s case, but to enable them to decide whether it had one. Given that Optiver already had that information, although not necessarily its full extent, Tibra was not required to produce its source code for preliminary discovery.

The Appeal

Optiver appealed Justice Tamberlin’s judgment saying that it was being deprived of the very information it needed to determine if it had a case. The court agreed with Optiver reversing Tamberlin’s decision. They found that Optiver did not have sufficient information to decide whether to sue Tibra. What they had was circumstantial evidence that gave them ‘reasonable cause’ to believe they might have such a case.

The email at the heart of Optiver’s case was sent from a Tibra director, Mr King on the last day of his employment by Optiver. That email listed ways in which the algorithm ‘F1’ could be improved. It was so significant that its contents were reproduced in the court’s judgment.

When the existence of that ‘one email’ was combined with the timing and circumstances of the departure of the employees and the setting up of the Tibra business, the court found that Optiver had “reasonable cause” to pursue a case against Tibra, and that preliminary discovery was a part of that. They also found that there was no prejudice to Tibra caused by Optiver’s delay and that inspection of their current source code was highly relevant to the question of whether or not Optiver should proceed. They were confident that the confidentiality regimes generally used in these cases would provide sufficient protection to Tibra.

In the course of its judgment, the full court made some comments on preliminary discovery. In particular, it said that the point of preliminary discovery was to allow a prospective plaintiff, in circumstances where they have a ‘reasonable belief ‘that they may have a case against a prospective respondent, to go fishing for more information to decide whether in fact he does have that case.

The High Court

Tibra then sought special leave to appeal to the high court to reverse the full court’s decision. That appeal was given short shrift by the high court who rejected the application after just 18 minutes of hearing time! The High Court does not like its time wasted.

After his judgment was overturned by the full court, the matter was brought back before Justice Tamberlin. His job now, as directed by the Full Court, was to make appropriate discovery orders for the source code. Those orders were handed down on 6 February, 2009 over two years after his initial judgment in this matter.

Optiver sought from Tibra all documents comprising or relating to “Optiver programs”. There were twelve such programs as defined by a series of functions. Tibra was to provide all instances of documents connected with these programs including all historical versions, either archived or deleted. It included source and object codes as well as logs and any other documents that referred to them.

For those unfamiliar with the terms source and object codes, source code is the human readable code that software programs are written in. The machine itself cannot execute source code so it is first compiled into object code i.e. a series of machine executable instructions. The term document itself is of course interpreted very widely to include any ‘record of information’.
Tibra provided evidence of its extensive software holdings. It claimed there were 10,000 versions of its automated trading system and calculated that to fully comply with the Optiver request would take up to 20,000 hours of work which they submitted would naturally cause severe interruption to their business.

Tibra’s legal team therefore sought to exclude object code altogether and limit access to source code to specific date versions on the basis of oppression.

Bearing in mind the appeal court’s comments on the relevance of different versions of the code the court rejected these submissions, ordering Tibra to provide all versions of both source and object codes from their ‘historical source repository’ as well as any other documents referring to them. As for Tibra’s confidentiality concerns, access to computer codes and any documents over which Tibra claimed confidentiality was only to be provided to Optiver’s solicitors and third party expert, but not to Optiver itself. The solicitors and expert first had to sign confidentiality undertakings. As a further safeguard, a copy of the expert’s report to Optiver was to be sent to Tibra, to ensure that it contained no confidential information

And Now?

About nine months have passed by since these orders were made. Clearly there was a lot of work to be done by Tibra to make discovery and the matter is reportedly back in court this month.
For now, barring further preliminary discovery stoushes, it is a matter of waiting to see if Optiver gets what it needs from the ‘crown jewels’ to take the next step.
This war is already almost three years old with countless skirmishes. Despite this, the outcome is far from over. Tibra strenuously denies the allegations and intends to continue to defend the matter vigorously.

It should be remembered that there is nothing to stop a former employee or employees from walking out of their jobs and subsequently competing with its ex-employer.
The issue is how it is done and whether any confidential information of the employer has been stolen and misused in the process.

It is therefore important for the former employer to come to court with some reasonable basis for their suspicions that this has happened. As this matter ably demonstrates, a single email can provide that basis. The fact that Tibra did not seek to explain away that email only adds to its potency.

This matter should demonstrate once and for all, the importance of retaining proper email records for a business. This must include both the full contents of the email along with details of when and how it was sent.

David McGrath is Director of e-Litigation Solutions. He currently provides independent e-discovery, information management and technology consulting services to corporates, law firms and the courts. David holds degrees in Law and Computer Science and post graduate qualifications in Technology Management.