Many FX trades are completed automatically using execution algos yet the ability to analyse those trades and identify when best to make them has been impossible. Yangling Li, Head of Quant Research at BestX, says FX traders need to have transparency over the timing of all trades – be they spot or algo – if they are to avoid information leakage or signalling risk.
“Signalling risk – or information leakage – is a problem for FX traders, particularly in markets where liquidity is reduced,” Mr Li says. “If firms can identify when trading volumes are at their highest, they can choose to execute trades at those times and then match internally. This avoids having to hedge FX risk with at an external venue which in turn limits signalling risk.”
Understanding algos
But identifying when most trades are being conducted using algorithms is easier said than done. Among the challenges is the time taken to conduct certain algo trades.
For example, time-weighted average price (TWAP) algorithms divide large orders into smaller ones which are then spaced out over time. The idea is to reduce market impact and may see a trade completed over many hours.
Mr Li says: “Finding the density of spot trades is very simple; every five minutes sum up the trades and then you check the average across the month which produces a nice curve. Algos are different as they can last for hours for example starting at 8am and ending at 4pm.”
The ability to understand TWAP trades are made even more complicated by versions that introduce randomisation in the timing of execution. For example, a 60-minute trade that has been set to trade once every minute, may be reconfigured to execute randomly to reduce the predictability and signalling from orders.
These algorithms may be yet further designed to execute more flexibly to achieve a better price, causing even greater deviation from an execution schedule, making it harder still to identify when trades are taking place.
Eight million trades
Given that BestX was born out of the demand for independent analysis to enable traders to ‘implement a flexible, rigorous, justifiable and repeatable execution process in accordance with market and regulatory evolution’, it is no surprise they had the technology and expertise to devise a solution.
Mr Li says BestX’s algo density metric provides FX traders with the information they need to identify when most algos are traded, even if they were executed over long time periods, by examining eight million trades in the company’s database. Starting on the 1 January 2021 and ending on 10 October 2021, based on the London time zone, BestX’s metric produced interesting results
Mr Li says: “We found that the algo volume profile is very different from the one we have observed for spot trading and that the peak algo density it as 4pm London time. We also found that the algo usage picks up smoothly at Asian and London time and then drops sharply at New York time.”
The algo density metric was also able to determine the times when different kinds of algo were traded most heavily.
Mr Li says the Interval algo proved popular during the European afternoon, while the Opportunistic algo has smoothed usage and is ‘the most popular across the board’. BestX also detected the algo density by currency.
“The G10 algo usage is smooth while emerging market algo usage jumped up at European afternoon,” Mr Li says.
Welcome development
The upshot of the advent of a fully functioning algo density metric is improved efficiency and heightened execution for FX traders. Mr Li says: “The metric has proved popular because the buyside can compare their own algo density against the overall market and understand why they might not be trading well, or if there is any room for improvement.”
And Mr Li believes the metric will influence how FX traders behave, ultimately determining when they trade and delivering best execution.
“If we think about the dark ages when there was no data and traders operated at random; there were huge spreads. As time has gone by, more market data has helped them trade at more liquid times and with better results. Now we have an algo density metric and even more information about what the market is doing which will have real benefits for best execution,” he says.