By Pete Eggleston, Co-Founder & Director and  Dean Markwick, ML Quantitative Researcher at BestX Ltd.
By Pete Eggleston, Co-Founder & Director and Dean Markwick, ML Quantitative Researcher at BestX Ltd.

Market Impact and Decay Around the WMR Fix

Market impact can mean many different things but in essence it is an estimate of the movement in the market price during and following a buy or sell trade. The act of buying pushes the price up and the act of selling causes the price to fall

Market impact can mean many different things but in essence it is an estimate of the movement in the market price during and following a buy or sell trade. The act of buying pushes the price up and the act of selling causes the price to fall. How much the price moves with each trade is known as market impact and is a key factor to any execution strategy. 

Measuring the market impact in FX is a difficult task as there is no central order book, tape or reporting authority, therefore we must be creative and one way to estimate this is to analyse the WMR fixing window to estimate both the amount of market impact a currency displays and how the market impact reverts after a trade has finished. The WMR fixing window provides an experiment to measure such a phenomena. The majority of trading participants in this window will be attempting to average in over the window to achieve as close to the fixing price as possible, therefore we can view the behaviour over the window as, for example, a single TWAP (time-weighted average price) trade. Each second of execution is displaying market impact and everyone executes their own orders pushing the price away from its origin at the start of the window. 

Once the window closes the impact decays, but not fully as there is a new price. In this article will be showing how the market impact looks for multiple currencies paying particular attention to the last quarter.

Calculating Market Impact

To characterise the market impact over the WMR window we use the following methodology:

  1. Snap the price at 15:57:30 - the start of the fixing window.
  2. Calculate the second by second return from the snapped price up until 16:05:30.
  3. Normalise the returns based on the price change from the start of the fix to the end of the fix (16:02:30).
    If the price is higher at the end of the fix, this was a positive day and the majority of people were buying.
    If the price is lower at the end of the fix, this was a negative day then the majority of people were selling.

When we aggregate over multiple days we don’t want positive days cancelling out with the negative days, hence why we must normalise the return based on whether the market was buying or selling on that day. We can then average the market impact on each day over the different quarter to judge how it has evolved.

There are two properties that we will observe over the window. Firstly, the price will be driven from the starting price due to the large amount of trading activity over the window, there will be an imbalance of buyers and sellers and the price will reflect this change. Secondly, the price will decay from its peak as the window finishes, this is because the amount of buyers and sellers returns to normal and finds a new equilibrium.

Here in Figure 1 we can see an actual example of this market impact. 

Fig 1 A real example of market impact of GBPUSD in Q2 2019

The increase from $t=0$ to $t=1$ shows the price of GBPUSD being driven by the flow of the fixing window. The price is driven on average 3.5 bps away from the starting price. It then reverts by 0.25 bps, leaving a new equilibrium price that is 3.25 bps away from where the price was at the start of the window. This characteristic shape across the fixing window highlights both market impact and the subsequent decay of the market impact. We now proceed to calculate the market impact across the fixing window for a number of different currencies and compare how it has changed across the quarters from 2019 to 2020. We then calculate the behaviour over the last day of the quarter and the last day of the month.


There has been recent discussion in the trade press around the behaviour of AUDUSD over the fixing window. This type of analysis can help shed some light on what has happened in the last quarter compared to the previous quarters.

Fig 2a AUDUSD market impact across different quarters

In Fig 2a we can see that AUDUSD has had a departure from normal behaviour, with an average of 6 bps of impact in Q1 of 2020 compared to the usual 3bps as seen in the previous quarters. However Fig 2b indicates how incredibly noisy this data is and how any conclusion drawn must be taken with the uncertainty of the observations under consideration.

Fig 2b Same graph but with the uncertainty of the measurements


For G10 currencies Figure 3 shows that most have seen a departure from the normal with greater market impact across the board.
Emerging Markets

Fig 3 G10 currencies

A similar pattern for EM currencies with an exception being USDTRY. So whilst the market impact is smaller Figure 4 shows 1bps of reversion in USDTRY last quarter.

Fig 4 Emerging market currenciess

The behaviour of EM currencies can be more difficult to estimate due to the overall lower use of the Fix compared to G10 currencies.

Scandinavian Markets

Figure 5 shows that the Scandinavian currencies have seen a large increase in market impact.

Fig 5 Scandinavian market currencies

Quarter End

We can also extended our analysis to assess the behaviour on quarter end dates. Here we look at all the quarter ends from 2017 to 2020 and compare them to normal days.

Fig 6 Quarter End for G10 currencies

However with only 4 quarters in a year, this makes it tricky to differentiate between signal and noise.


Month End

We can include all month end dates.

Figure 7 is less noisy than Fig 6 and it appears that there is a constant shift upwards in the market impact compared to a normal day. 

Fig 7 Month End for G10 currencies

This highlights the potential cost of executing on month end over the fix, the price is more likely to be driven away from the starting fix price.


To show that this effect is real we repeat the procedure but for a window that is 5 hours earlier (11 am) in Q1 2020.

Fig 8 Comparing the fixing window to an 11am window

For all currencies (except USDMXN) Figure 8 shows that the market impact is larger in the fixing window and there is a more noticeable decay after the window closes. 


This shows the effect of all participants executing in the fixing window compared to a random window where it is unlikely that everyone is executing and the volumes being executed will be orders of magnitude smaller. The difference of USDMXN can be attributed to the lower usage of the fix as mentioned previously.

When we focus on AUDUSD in the last two quarters we find a similar pattern.

Figure 9 shows that the overall impact in both the control window and the fix window is larger in 2020 Q1 compared to 2019 Q4 which can be attributed to increased volatility. However, the fixing window always has a larger market impact that decays which is dominated by the market impact of trades targeting the fix.

Fig 9 Comparing the fixing window to an 11am window for AUDUSD in the last two quarters


Overall we have shown how market impact can be calculated and visualised. By assessing it over a number of different currencies we can see that it varies from quarter to quarter and there has been a large increase in market impact over the last quarter. Furthermore, there is a larger market impact at both quarter end and month end. This is as expected given the increase in volatility, and volume of rebalancing trades due to large movements in indices, and doesn’t necessarily indicate any nefarious behaviour around the Fix. Market impact alone is an inherently difficult problem as the amount of noise present in the observations can easily wash out any apparent signal in the data. In the Figures where the error in observation is highlighted (Figs 2b, 6 and 7) the large scale of uncertainty must be taken into account before drawing any kind sweeping conclusion. There is evidence to support the theory that there is more permanent market impact (i.e. there is a smaller decay in the price) created at month ends compared to `normal’ days. This evidence would therefore suggest that, where possible due to mandate or operational constraints, it may be more optimal to rebalance on days other than month-ends.