Thomson Reuters to trial randomization on FX matching platform
A new concept of holding orders to tackle latency arbitrage will be trialled for pauses of at least three milliseconds on USD/MXN trades on Thomson Reuters Matching, just weeks after rival EBS completed the roll-out of its latency floor.
Thomson Reuters is to begin randomizing execution of orders on its flagship foreign-exchange matching platform as it seeks to rebalance the system and promote equality of treatment for users.
The vendor plans to trial the process – which will be known as randomization of order processing – with a single currency pair, likely to be USD/MXN, and will then make a decision on whether to extend it to other currency pairs.
The move was unveiled on Tuesday as part of a broader swath of rulebook changes intended to improve the behaviour of users of the platform, which has been dogged by the rise of high-speed traders in recent years.
“The purpose of this randomization feature is to make the matching market more inclusive by establishing a minimum response time below which market participants will gain little or no additional advantage when making or taking prices very quickly,” says Thomson Reuters in a client guide to the proposed rules seen by Euromoney.
Randomization is a central component of the new ParFX platform, which launched in April 2013 with the backing of some of the largest banks in the FX market, such as Deutsche Bank, UBS and Barclays. ParFX applies a randomized pause of between 20 and 80 milliseconds to all order elements to engender a level playing field for users.
Meanwhile, EBS last year began trialling its latency floor, which introduces a random-length batching window of one, two or three milliseconds, in which all messages – including bids, offers, buys, sells or cancels – arriving from an institution are randomly assigned to a unique row per institution within a queue set.
When the batching window completes, all messages are released to the matching engine in a round-robin fashion.
Thomson Reuters has not yet determined the exact calibration of its planned randomization logic, but on the basis of “extensive analysis of behaviour and trade data”, it plans to randomize processing for at least three milliseconds during the pilot phase with USD/MXN.
Crucially, the platform plans only to apply randomization to orders that aggress the market. “Maker orders that do not match on entry [ie that do not ‘cross the book’] will not be subject to the randomization and will continue to be processed without delay,” the draft rules state.
The news comes just weeks after EBS completed the roll-out of its latency floor, which had been trialled on AUD/USD since August and USD/CHF since September. After an extended period of analysis and testing, the latency floor was extended to a further eight pairs on February 3 before being finally rolled out to the platform’s two most liquid pairs – USD/JPY and EUR/USD – on February 17 and March 3 respectively.
“The latency floor addresses the ‘technology arms race to the bottom’ whereby continuous incremental spend on the fastest technology is required in order to stay competitive,” says John Schoen, co-head of EBS Market. “It is a way of ensuring that speed as a standalone strategy is not a prerequisite for success on EBS.”
After a change in the granularity of pricing in 2012 that saw EBS move from quoting in tenths-of-a-pip to half pips and full pips, the latency floor was the second part of its push to improve the trading experience of its users.
However, it was only intended to address the race for speed that takes place in microseconds and single-digit milliseconds, which means some users that trade at a comparatively slower pace might not have noticed much change.
“What we have really addressed is a small sub-section of our participant base for whom microsecond performance differentials created an unfair advantage, but that is no longer the case,” says Nichola Hunter, co-head of EBS Market.
“We see periods of competition, when more than one institution competes within a batching window, occur roughly 10% of the time throughout the trading day.”
Other planned changes on Thomson Reuters Matching include a tightening of the minimum quote life (MQL) component, which is designed to ensure all orders are shown in the market for long enough to give users a fair chance of trading them.
MQL had previously been imposed on most instruments for a period of 250, 500 or 1,000 milliseconds, but Thomson Reuters is now planning to introduce an MQL of 100 milliseconds for pairs that include the euro, yen, Russian rouble and Swiss franc.
As for pricing of currencies, Thomson Reuters last year reduced the granularity in which three of its core currencies – the Mexican peso, South African rand and the rouble – were quoted, increasing the tick size from one to five pips. The move was intended to combat disruptive behaviour, such as price flashing, and create more stable prices on the platform, but further changes to tick sizes are not planned at this point.
The draft rules are open for consultation until April 30, and Thomson Reuters intends to publish the final rules by June. However, it will proceed at a much slower pace on randomization, with final decisions on whether to extend the measure to other currencies not expected until next year.
“Through a combination of platform controls and behavioural rules, the new updated rulebook will seek to enhance liquidity for all clients trading on Thomson Reuters Matching,” says Phil Weisberg, global head of FX at Thomson Reuters and former chief executive of FXall, which the vendor acquired in 2012.
“We take our responsibility to maintain a fair and orderly trading experience for all spot matching market participants very seriously.”