Is a safe algo also a stale algo?

Regulators are keen to ensure they can manage the risks associated with algorithmic trading but the extent to which they do this will determine the impact of this on execution performance.
By None

It's understandable, especially post-flash crash that regulators want to ensure algorithms do not pose a systemic risk to financial markets, but could the imposition of checks undermine execution performance?

This all depends on how deeply regulators want to analyse algos, but many argue that no tangible detrimental effect on execution quality is likely.

As discussed last week, there are various proposals to vet algos across the globe. Some are focused on pre-trade risk controls while others require explanations of how individual algos function.

After the 6 May 2010 ”flash crash' in the US, more stringent pre-trade risk checks seemed a sensible measure to prevent a similar occurrence. Under a new rule that was passed by the Securities and Exchange Commission in November 2010, brokers are unable to give their clients direct, unfiltered access to trading venues and are required to put appropriate risk management and supervisory procedures in place to prevent erroneous orders. Such extra checks are not seen as having any significant bearing on the cost of a trade for the typical long-only asset manager. Similarly, simply explaining the design and functioning of an algo to regulators – currently required in Thailand, where algos are approved by the stock exchange in a relatively short timeframe – would have little impact on execution performance.

So does this mean market participants shouldn't be worried?

The primary reason that vetting algos may have a negative effect on execution performance is the potential delay in time to market for new algos or enhancements or customisations to existing ones.

If algo approval becomes a regulatory requirement in Europe, which has dozens of algo providers, and the authorities mandate an examination of the underlying code used for each, the process would be hugely time consuming.

The risk here is that algo strategies may become stale by the time they have been blessed and therefore risk failing to deliver optimal execution performance – a problem that has been emphasised by some brokers looking to launch algorithmic trading services in Indonesia, where prior approval is required.

A prolonged approval process could also limited the ability of brokers to supply customised algorithmic solutions for clients that wish to adapt to changes in liquidity conditions in a timely manner.

The trend towards customised algorithms is growing in both the US and Europe, particularly following the official launch of FIX Protocol's algorithmic trading definition language in March 2010, which speeds up algo deployment by allowing providers of algorithms to send algo specifications via vendors to buy-side firms using an industry-standard, computer-readable XML file.

What is the likelihood of regulators wanting or having to conduct a thorough examination of each algo?

By all accounts, most regulators seem unwilling to study the lines of code that make up an algorithm. Developing and retraining the expertise required to analyse algorithms in a detailed way is probably a challenge most regulators could do without right now. In addition, some market participants are quick to point out that handing this level of responsibility to regulators could mean brokers feel they are absolved from their own responsibilities.

Tim Rowe, manager, trading platforms and settlement policy, markets division for UK regulator the Financial Services Authority, echoed these concerns at the recent TradeTech conference in London, and urged new rules to focus instead on ensuring algorithms are tested adequately by brokers, with extra checks conducted at the trading venue level.

While the European Commission raised the possibility that “competent authorities” may have to sanction algorithms before they are used in its recent MiFID II consultation paper, some market participants have noted that extrapolating this to mean regulators will analyse algo code is far-fetched.

Another case in point is India, where the country's largest bourse, the National Stock Exchange (NSE), has been criticised for taking too long with its algo approval process. However, the NSE now seems to be relaxing its position after a recent circular permitted minor changes to be made to an algorithm without prior consent.

Click here to vote in this month's poll.