What role can FX algorithms play in helping to manage fixing orders?

June 2023 in Previous Features

Paul Aston looks at how when seeking liquidity around the WM/Reuters 4PM London Fix, currency algorithms can help. Investors can take better control of their trading, reduce costs, obtain market transparency and ensure best execution.

By Paul Aston, Head of Quantitative and Algorithmic Solutions at TD Securities

When seeking liquidity around the WM/Reuters 4PM London Fix, currency algorithms can help. Investors can take better control of their trading, reduce costs, obtain market transparency and ensure best execution. But in order to best utilize algorithms around the WM Fix, investors need to fully understand the mechanics of the WM Fix calculation methodology and how a particular algorithm works to achieve specific goals.

There are certainly new challenges to seeking liquidity around the WM Fix. Earlier this year, after lengthy investigations by several global regulatory authorities, it was determined that over time several major dealers in global foreign exchange had colluded to manipulate market pricing during the WM Fix calculation window. Most of the dealers involved eventually plead guilty to these charges and paid significant fines and penalties.

“…the first role that algorithms can play for investment managers seeking liquidity at the WM Fix is market transparency.”

In response, dealers industry wide began to institute changes to their compliance and operating procedures to ensure that client information during the fixing window is safeguarded and that fixing orders are filled and handled properly. In addition, the WM Company made proactive changes to its calculation methodology in order to make it less susceptible to market manipulation (1), the most important of which was to widen the calculation window from the WM Fix to 5-minutes from 1-minute. (2)

The combination of these developments significantly increased the market making cost of procuring liquidity at the WM Fix. As a result, many banks now charge fees to fill orders for WM liquidity. This has had an important impact on buy-side firms that have previously been able to obtain the WM Fix Mid-Rate at no apparent cost.


Now that transaction costs are an explicit part of managing fixing orders, buy-side firms must become far more active in their pursuit of best execution surrounding the WM Fix. This is where algorithms can add considerable value.

First and foremost, what many people do not realize is that the WM/Reuters 4PM London Fix does not represent a bona fide price that can be traded or replicated in the open market; it can only be negotiated for delivery between two counterparties. This is because the WM calculation methodology involves certain procedures that ultimately deviate from the results one would obtain when actively trading in the market. As such, the WM Fix should be really regarded more as a reference price index that summarizes the level of exchange rates observed over a particular window of time each day rather than as a dealable price that was actually attainable in the market.

Thus, in order to best utilize algorithms around the WM Fix it is imperative that one fully understands the WM calculation methodology in detail. (3)

In short, the WM calculation methodology may be thought of as a type of algorithm itself, which can be broken down and summarized into four distinct stages or processes. (4)

1. Data sourcing

2. Data sampling

3. Data completion and validation

4. Calculation of the fixing rate

The data sourcing and sampling stages are perhaps the most straightforward and algorithmically replicable parts of the calculation methodology.

In these stages, exchange rate data is sampled every second over a 5-minute fixing window, commencing 2.5 minutes before and ending 2.5 minutes after 16:00 London time. The data is culled from three electronic dealing platforms: Thomson Reuters Dealing 3000, EBS and Currenex.

Algorithmically, these stages replicate the prices one might obtain if they implemented a 301-slice, 1-slice-per-sceond, time-weighted average price (TWAP) algorithm beginning at 15:57:30 London time while connecting to the Reuters, EBS and/or Currenex liquidity platforms.

If the WM calculation methodology simply took the average of the sampled data, such a TWAP would have a pretty decent chance of replicating the WM Fix with minimal tracking error. But this is not the case. In the calculation methodology some additional proprietary steps are taken to clean the data and it is these procedures that transform the WM Fix into a non-dealable price.


Once captured, the exchange rate data is then “completed” and validated upon the conclusion of the WM fixing window.

Data completion essentially involves the process of ensuring that captured trade rates, i.e. – paid or given transactions, are paired with a complementary bid and offer rate, respectively. This involves computing the complementary side of the market by subtracting or adding a spread to the paid or given rate, respectively. The spread is obtained from the indicative bid-offer order rates that were obtained at the same instant as the trade rate.

We assume that this step serves the purpose of ensuring the completeness of the market information captured and to ensure that bids and offers have comparable sample sizes for statistical estimation purposes. However, it has the effect of injecting sample bias using pseudo exchange rates that were neither traded nor indicated during the fixing window, which will tend to push the WM prices away from any actual trading result.

Once this step is completed and validated, all trade rates that have been captured are pooled and segregated into bids and offers and from these resulting pools a median Trade Bid and a median Trade Offer are calculated independently. The official WM Mid Price is then calculated as the median of these two rates.

This last step is effectively makes the WM Fix non-replicable by any type of active trading strategy. This is because when one actively trades, the effective price that one obtains from a set of transactions is the arithmetic weighted-average of all executed prices; not the median of those prices.

Thus, when a dealer commits to delivering the WM Mid-Rate to a client they are essentially taking on active market making risk because the actual market cost of the inventory required, which must be acquired at the bid or offer, is likely to be more expensive than the WM Fixing price they promised to deliver. This is why banks have begun to charge fees for procuring WM liquidity.

When the WM Company widened the calculation window from 1-minute to 5-minutes, WM market making risk statistically went up by a factor of at least 2.24x (since volatility increase with the square root of time).

The WM/Reuters 4PM London Fix does not represent a bona fide price that can be traded or replicated in the open market


In light of these facts, the first role that algorithms can play for investment managers seeking liquidity at the WM Fix is market transparency.

The simple 301-1 TWAP strategy highlighted above provides a base line for estimating the fair cost of acquiring WM liquidity and delivering it at the WM Mid-Rate. On any given day, simply take the difference between the WM Mid-Rate and the 301-1 TWAP price implemented at bid or offer, accordingly.

In theory, there is an upper-bound for this cost, which is given by the Chebyshev-Cantelli inequality from statistics. It turns out that this cost cannot be greater than the standard deviation of the observed pricing data sample observed over the fixing window, plus one-half the official WM bid-offer spread (the difference between the median Trade Offer and median Trade Bid). This suggests that the cost of procuring WM liquidity is effectively proportionate to volatility.

Figure 1 shows an estimate of the difference, in pips, between the WM Mid- Rate and a 301-1 TWAP price for EURUSD over the past month.

What we have been seeing in practice is that many banks are indeed implementing a simple 301-1 TWAP strategy, or variations thereof, to process fixing orders under their new compliance and operating protocols. Thus, fee structures across the street should be in line with these estimates with perhaps a premium added to account for the vol-of-vol.

By using a similar 301-1 TWAP algorithm, buy side firms can establish an on-going baseline for assessing the relative value of different fee structures when implementing fixing orders on different platforms.

There is no need to implement this algorithm live to obtain useful results for this purpose, but for those buy-side firms willing and able to take full control of sourcing fixing liquidity implementing such an algorithm can help ensure that their costs never exceed the WM Mid/301-1 TWAP deviation on any given day.

Implementing this algorithm can be achieved in-house, if ones systems are set up appropriately or it could be accessed on any number of platforms that provide flexible and transparent functionality to configure algorithms. When accessing such algorithms one should keep in mind that such platforms may charge fees to use their algos and/or agency fees to access liquidity. Hence, one should always be diligent to always perform transaction cost analyses on the liquidity they receive in order to fully assess the relative value of dealing on various platforms and accessing value-added trading technologies.

Going beyond a simple 301-1 TWAP, investors may want to consider designing or using more advanced algorithms that might reduce the costs of accessing WM liquidity even further. Such algorithms might incorporate smart order routing between dealing platforms; a greater diversity of liquidity provides in addition to Reuters, EBS and Currenex; logic to help reduce market impact and information leakage; and/or dynamic trading capabilities coupled with real-time replication of the WM calculation methodology to optimally vary execution sizes and timing to minimize the error between the effective algo price and the resulting WM Mid-Rate on a given day.

Investment managers should also keep in mind that algorithms should probably be combined with other more traditional methods of obtaining WM liquidity from dealers and/or combined with many of the emergent agency platforms that have been specifically designed to aggregate and net fixing liquidity for multiple market participants. These latter types of platforms not only provide the benefits of netting with other counterparties, which helps to reduce over volatility, impact and potential market manipulation, but can also provide additional transparency into the microstructure and depth of liquidity surrounding the WM Fix.

As always, with any platform solution and product offering, buy-side firms should keep an eye on fees and take an increasingly assertive approach to trading and assembling information to ensure best execution. In this regard, algorithms can not only help, they are essential.

1. For a comprehensive overview see documents and press releases published by the WM Company at http://www.wmcompany.com/wmr/index.htm

2. See The WM Company (2014h) “WM/Reuters Spot Rate Service – Methodology Changes”, November 14 http://www.wmcompany.com/pdfs/WMReuters_Spot_Rate_Service_-_Methodology_Changes.pdf

3. For full documentation of the calculation methodology see WM/Reuters “Spot & Forward Rates Methodology Guide”, http://www.wmcompany.com/pdfs/WMReutersMethodology.pdf

4. This breakdown of the WM calculation methodology is an analytic overview produced by the author at TD Securities, the details of which may be found in Aston, P. (2015) “Estimating the Cost of Procuring WM Liquidity. A Practical Approach”, TD Securities FX Quantitative and Algorithmic Solutions (QAS) Group white paper, February 16.