The half speed method: optimising algo timing on an execution desk

Originally published here: https://www.fx-markets.com/trading/7925081/the-half-speed-method-optimising-algo-timing-on-an-execution-desk

---

There is no single answer to “How long does it take to trade 100mio?”

As we demonstrated in a recent note the time taken varies according to:

• The currency pair e.g. EURUSD trades at least 10X more daily volume than EURSEK;
• The time of day e.g. USDMXN trades at least 10X more in peak NY hours than out-of-hours;
• The day itself e.g. volumes on a busy day might be 3x higher than on a bank holiday.

Execution algorithms are calibrated to make these adjustments on behalf of their users. For example, they might dynamically adjust to current conditions and trade faster in core hours than out-of-hours.

Traders decide on cost vs speed trade-offs

The classic trade-off for an execution trader is between cost and speed. This informs which algorithm they pick for a particular job.

At one extreme there is full risk transfer. The whole balance is executed instantly with no time for the market to drift. However, the cost of this is a full spread.

Alternatively, a patient algorithm will aim to save on spread cost but in return introduces more variance: the market has more time to drift away from the initial price before the order is filled.

Choosing a point on this continuum is the job of the execution trader. This will depend in part on the expectations of the portfolio manager and the nature of the underlying flow.

Examples

Below is an illustrative example of a fast execution that worked well. The trader bought aggressively (grey zone) following which the market kept rising. If they had traded more slowly they’d have paid a higher price.

Illustrative image only

One explanation for this could be that the flow itself had short-term alpha. This means that the portfolio manager correctly anticipated that the market would re-price higher and it shortly did. Trading fast was justified.

Below is an illustrative example of a fast execution that didn’t pay off. The trader sold aggressively and then the market mean-reverted. If they had traded more slowly they’d have sold at a higher price overall.

Illustrative image only

One explanation for this could be that the order over-traded and caused its own market impact. When it stopped the market mean-reverted. There was no need to trade fast since the flow apparently didn’t have short-term alpha.

Evaluating your own flow

Knowing the optimal speed of execution therefore depends on understanding the short-term alpha characteristics of your own flow. A simple but effective way to do this is to look at the half-speed TWAP metric.

For example, let’s say you sold 100m USDCAD between 12:15 and 12:30. This execution happened during the shaded grey area and the all-in rate is shown with the blue dotted line. The half-speed TWAP metric assumes you trade half as fast - i.e. the execution takes place between 12:15 and 12:45 instead.

Illustrative image only

This is represented on the chart as a combination of the grey and blue shaded areas. It calculates a simple TWAP execution rate for this 30-minute period, assuming it hits mid-price every one second. The execution rate is represented on the chart with the grey dotted line.

Now you compare the two. In this case the execution was inferior to the half-speed TWAP. We can see on the chart it looks like the order has over-traded. After it stopped the market mean-reverted higher as we see in the blue shaded area.

Using your own data to improve performance

Looking at each order will be noisy due to market drift. It is more meaningful to aggregate all your orders over the last year. As an illustrative example you may see the below.

Illustrative image only

Across the 347 orders we can see there appears to be a tendency to over-trade:

• The average execution is inferior to the half-speed TWAP by 2.1 basis points, which is sizeable in the context of an overall average cost of 3.6 basis points;
• Moreover 66% of orders lost to the half-speed TWAP.

In this illustrative example, the savings to the firm could be significant. 2.1 basis points times $34.7bn of flow equates to over $7m of potential savings annually.

Having hard data is crucial as it justifies updating the trading policy and taking more time risk on individual executions. It can help produce a business case for giving the trading desk more time to work orders and add value.

The nice thing about this metric is that it speaks specifically to each firm’s own alpha and orders. The results won’t be the same for everyone. Analysing the data and adjusting behaviour might offer a meaningful way for execution desk traders to add value.

---

The views expressed in this article are the author’s personal views and should not be attributed to any other person, including that of their employer.