SIREN FX unveils new approach to measuring algo performance

TWAPs persist as the FX market’s preferred choice of execution algo, despite often being the worst choice of algo strategy for buyside clients, according to Walton. One reason for the popularity of TWAPs in the market is that they are measurable, he adds, providing not only algo execution but also a way to benchmark algo performance to what the TWAP price would have been. “Measuring FX algorithms can be tricky. If the goal is to hit the TWAP price, that entails trading linearly and periodically at the same time, which is the worst way to execute in terms of market impact,” Walton says.

In addition, he warns that because most passive TWAP algos are no longer vanilla, they can easily move off their trajectory of accumulated positions. “Normally a trajectory of accumulated positions is linear, but a passive TWAP is allowed to deviate from that,” says Walton. “If clients are then still measuring performance with a TWAP algo assuming that linear performance, that measure of algo performance will be inconsistent.” 

Recognising limitations 

This inconsistency is also evident in other benchmarks used in FX, most notably the WMR 4pm Fix, which is essentially a five minute TWAP itself. “Yet we know that while a TWAP is the easiest to detect and measure, it is also means the wider market can detect and trade against that activity,” Walton explains. “The limitations of executing around the Fix have been recognised for over a decade. It creates this recognisable V-shape in flow, which stems from the market being able to read those TWAP executions that take place during that window,” he adds.

The SIREN FX benchmark was established following a 2018 meeting Walton had with the Bank of England about improving outcomes around the Fix. This led to the development of the benchmark as an alternative to TWAPs, achieved by using a different algo strategy – essentially an implementation shortfall algo which is used in reverse, Walton explains. “TWAP is not a great strategy to use if you care about price,” he adds. While not so well know in FX, in equities there is a market-on-close (MOC) algo, which aims to get close to the Fix but also to reduce market impact. 

The SIREN FX benchmark is based on the same principles of calculating the rate while also providing optimal execution for the buyside, trading less, reducing market impact but still increasing as it gets close to the Fix, according to Walton. He adds: “It is more sophisticated than a TWAP, which is only for executing and has no way to target the price.” Following FCA authorisation in 2019 many leading FX providers now allow for execution against the SIREN FX benchmark, including Goldman Sachs and NatWest.

Optimal execution performance

More recently, SIREN FX developed the SIREN FX Algorithm Reference Rate (SARR) using the same measurement for the Siren benchmark to calculate the performance of any benchmark execution using an FX algo. This new reference rate extends the curve to the length of the algo window and then calculates, using the SIREN FX benchmark methodology, what would be the optimal execution rate for that algo. “The reference rate is essentially a more advanced form of TCA analysis for algos,” says Walton. “It provides a new, entirely independent way for banks and the buyside to measure optimal execution for any algo. This provides the market with a more accurate and standardised representation of algo execution performance than can be achieved using a TWAP.” 

Standard TCA metrics will compare the price of the child orders using TWAP, while the new measure compares the performance against optimised execution, based on optimal execution frameworks, including the original Almgren–Chriss model, long considered the foundation of market impact modelling in equities. 

“If a bank measures its own algos, there will always be the question of whether they’re marking their own homework. As a third party, we can apply an established, academically recognised model independently,” Walton adds. “Too many buyside firms have been missing out on the advances in the development of execution algos due to a reliance on ineffective TCA. Out reference rate provides a new innovation to the space which can provide the market with a far more effective measure of algo performance.”