Despite an official report into the market crash of 6 May, its causes are still being hotly debated by US market participants. A crucial part of the puzzle is a large sell trade of E-mini futures contracts by buy-side firm Waddell and Reed.
A report into the ”flash crash' issued on 1 October by the Securities and Exchanges Commission (SEC) and the Commodity and Futures Trading Commission (CFTC) said the trade was executed using a participation algorithm that only took into account volume as a parameter and sold a lot of contracts very quickly as a result of high trading levels.
Alison Crosthwait, director of global trading research at agency broker Instinet, noted in her report on the crash that a “price limit provides an important layer of intelligence and protection” for users of algorithms. Regulators say no price parameter was in place but, Eric Scott Hunsader, founder of data analysis firm Nanex, disagrees: “The algorithm used definitely takes price into consideration.” Hunsader says that Waddell & Reed is flying the developer of the algorithm to its offices on 19 October to lay the issue to rest. When contacted by theTRADEnews.com, a spokesperson for Waddell and Reed was unable to confirm whether a meeting was scheduled to take place.
If the algorithm is found to be price sensitive, contrary to the case made in the SEC/CFTC report, it would suggest that the trade had insufficient unusual features – apart from its 75,000-contract size – to be legitimately identified as the primary ”trigger' for the crash on a trading day that exhibited many exceptional characteristics .
However, sources say, the apparent need to interview the algorithm's developer – and Waddell's continued silence – raises more questions.
Currently even the source of the algorithm is unclear: it may have been supplied by the executing broker or a third-party provider. Was the algorithm built with price sensitivity as a parameter? Was that parameter overridden or otherwise factored out? When the effect of the algorithm was seen why did traders not intervene? Why has Waddell and Reed, a Kansas-based mutual fund with $60 billion in assets under management and 99 years investment experience accumulated by its top three staff, not provided a statement on its version of events other than assert “The events of May 6 involve multiple issues that transcend the actions of any single market participant”?
On the first point it would seem unlikely, but not impossible, that price was not among the parameters included at the algorithm's construction. “One of the things you invest a lot of time into, when you are building an algorithm, is getting the parameters right,” says Chris McConville, executive director for direct execution services at broker UBS. “It's more data intensive but you need to have parameters at a stock level.” he adds. Either volume or price must be taken into account when building an algorithm says McConville, but setting an algorithm to respond to changes in the wider market rather than a specific instrument can cause malfunction. For less liquid instruments failsafe alerts would normally be triggered to warn the trader of possible market impact, however for very liquid, high volume instruments such as E-mini futures contracts, these may not have been necessary.
Regardless of the algorithm's design, it would seem that a number of in-trade checks that are widely considered normal market practice were omitted on 6 May. Given that the algorithm executed the same number of trades in 20 minutes that it might normally execute in five hours, it is possible that the trader did not fully consider market conditions and used too aggressive a setting, i.e. asking the algorithm to chase volume rather than passively placing orders on a book and waiting for them to get filled. It is unclear too whether the trader was able to change aggression settings midstream. Even if the algorithm is built with a price parameter set as its default position, algorithms can be customised and if a trader misread the signals in the market he could have moved the goalposts too far. However algorithms can also be built in a simplistic way e.g. I want to sell nine of every 100 contracts sold in the market therefore I set the algo to trade sell 9% of the market volume.
Whatever the arrangements for setting it up, both the buy-side firm and its executing broker would normally deploy members of staff to oversee the trade and react to unusual market conditions.
“You can bring in checks and measures but at the end of the day if a trader doesn't know how to use an algorithm properly, he can trade in this environment in a way that will cause this [crash] to happen,” says Steve Wood, founder of consultancy Global Buy Side Trading Consultants. “With multi-broker broker-neutral platforms you could have access to 50 brokers with ten algorithms, so your buy-side trader's expertise is knowing which algorithm to use,” he concludes.
In addition, trading system vendors have confirmed that most execution management systems used by buy-side firms would be able to provide an audit trail for any equity market transaction, identifying the algorithm and the parameters that were applied.
If it is found that the algorithm was price sensitive and that it functioned as intended, the official cause of the crash will have to be readdressed, with the inability of market participants and mechanisms to absorb massive selling pressure likely to gain greater prominence.
Hunsader asserts that highlighting the Waddell and Reed trade as a trigger seems unwarranted if it did not significantly pressure the market. By contrast he says, high-frequency trading firms were dumping E-mini futures contracts to avoid accumulating a position. “I read the [SEC/CFTC] report and it dawned on me that when it discusses the HFT firms becoming net sellers whenever they reached 2-3000 contracts, it discusses their trading as though it were normal. They dumped them as fast as they could without regard to price. [The report] should have at least mentioned this!” he says.