FX: Randomization gains favour, despite flaws
FX market players are gaining confidence in randomization as concerns over the potential for wider spreads and reduced certainty of execution have been allayed by market experience.
While it is an effective means of excluding pure latency arbitrageurs, randomization – the deliberate addition of delays to messages received from FX market participants – is not without its drawbacks. For example, there is the risk that some ‘friendly flow’ receive fills that are slightly impaired compared to a strict first-in-first-out model.
Roger Rutherford, ParFX
However, ParFX chief operating officer, Roger Rutherford, reckons randomized matching technology has proven itself to be an effective mechanism for deterring disruptive trading behaviour in the FX market, nullifying the speed advantages of dark fibre, cross connect and low latency systems and models.
“Our pause is meaningful enough to nullify disruptive traders whose strategy relies solely on speed, but meaningless to those that have a genuine trading need, who are looking for firm, executable liquidity and who compete on strategy,” he says.
A price maker attempting to cancel whilst takers attempt to take the price is the most relevant scenario for reducing latency arbitrage, according to Alexis Atkinson, head of order driven markets at EBS.
“Our latency floor mechanism has enabled a more diverse range of participants to make markets successfully,” he says. “Enabling a wider range of participants to be successfully quoting builds a stronger market, improves price discovery and reduces maker concentration risk, which is particularly important in stressed market conditions.”
Thomson Reuters lagged other platforms in introducing randomization, a delay that head of FX trading venues, Paul Clarke, attributes to the need to develop the right approach to adding delays and reordering messages.
Negative outcomes associated with some other approaches may include inadvertently continuing to provide a significant statistical advantage to participants who can respond more quickly than the delay period, he explains. “Our mechanism should lead to more equitable outcomes in price making and price taking for market participants on venues on which it is deployed.”
According to Clarke, randomization has helped to reduce spreads because the design provides some protection to market makers to encourage market-making activities. It has also been a key enabler for Thomson Reuters’ move to more frequent market data via the binary multicast feed.
Sceptics have described randomization as a short-term solution to technology deficiencies that merely leads to wider spreads and reduced certainty of execution. But Curtis Pfeiffer, chief business officer at Pragma Securities, says it is his firm’s experience (and it is conducting more quantitative research on the topic) that the certainty of execution may even be increased as a result of randomized matching because there will be less – possibly up to 50% less – quote fading.
Rutherford agrees that describing randomization as a short-term fix is unfair and says there is no evidence of it causing reduced certainty of execution on his platform.
“If anything, the stability and firmness of prices increases execution efficiency, as participants are less likely to be pipped by disruptive trading strategies,” he says. “Randomization has therefore proven itself to be a legitimate and meaningful solution to disruptive low latency trading strategies and successfully creates a market of genuine interest, reliable price discovery and firm liquidity.”
Improved market surveillance also has a role to play in ensuring fair and effective markets, but as a complement to randomization rather than an alternative, according to Clarke. “Market surveillance is essential for identifying behaviours that may not help provide an orderly market, or indicate suspicious activity, but would have no impact on reducing the extent to which a market participant’s speed determines their success in trading activities on matching.”
Atkinson observes that market surveillance addresses different issues to randomization. “With market surveillance, you can observe the effects of a fastest-takes-all setup, but that is different to implementing mechanisms to reduce the magnitude of a speed advantage,” he says.
Rutherford, meanwhile, refers to the timeliness factor. “Market surveillance only identifies disruptive and disorderly behaviour once it has occurred – it doesn't stop it from occurring in the first place,” he says. “Furthermore, the effectiveness of surveillance tools depends on their ability to identify these behaviours and also whether the punitive measures are an effective deterrent.”
One of the virtues of the lightly regulated FX market is that different models can exist side-by-side, concludes MarketFactory executive chairman, James Sinclair. “There are times and currency pairs when a trader may find a randomized pool gives better execution quality and other times when a strict first-in-first-out model produces better results. Different pools may also each have their time and place for a given trader, depending on the range of randomization times and whether cancels are randomized or not.”