Order stitching can play a significant role in optimizing trading strategies and evaluating their performance accurately. In this blog post, we'll delve into what order stitching is and why it’s important, as well as the nuances associated with the calculation of analytics in TCA and how we handle them in our own TCA.

What is order stitching?

Some traders send orders once and let the order finish trading (or not finish trading) according to the execution algorithm’s trade plan. Other traders divide a single order into multiple waves, sending multiple orders on the same date, symbol, and side throughout the day, as shown in Figure 1 below. Such waves may have periods of pause in between or they may overlap. In each successive wave, a trader may change the trading strategy–switching from a liquidity-seeking algorithm to a VWAP, for example–or update parameters like limit prices, volume caps, and more. Managing an order in waves aims to optimize execution strategy, and traders need access to meaningful cost measurement tools that can help determine the effectiveness.

Order stitching is a methodology for viewing waves as a single order in transaction cost analysis, which can help a trader evaluate execution performance more robustly. If the trader reviews performance of trades without stitching waves together, they will be treating the individual waves as distinct though they are part of a single parent order and the performance of earlier waves is likely to affect the performance of later waves.

For example, in the image above, there are three waves traded throughout the day. In the first wave, the order trades over a period when the price of the instrument is increasing. Because the order is a buy order, it is possible and perhaps probable that the order itself is contributing to the price change by way of market impact. In wave 3, we see that the price isn’t moving very much by comparison. This example illustrates a common outcome–the first wave of an order, just as the initial portion of any single order, is likely to create more market impact than consecutive waves. If a trader does not incorporate the dependence of the performance of the waves into their performance assessment, they could be underestimating the impact they’ve created over the course of the full parent order and missing out on the broader view of their trading performance. Combining the orders gives a measure of cumulative cost from all order slices versus the initial arrival price.

Looking at the individual waves evaluates the performance of the execution algorithm itself–was it able to access passive or dark liquidity, for example? Could it earn more spread than it paid? Did it create unnecessary market impact? Aggregating the orders to treat them as a single parent order and understanding the performance of the order’s execution as a whole will help to evaluate the real cost of trading that order–important for portfolio optimization–and help evaluate whether the trading strategy, including dividing into waves, was successful overall.

How does stitching work?

Stitching waves into a single order has simple elements and more complex ones. The simplest measurement to start with is the fill quantity. Consider the three waves depicted in the image above. If we combine the waves together into a single order, we can simply sum their total executed quantity to derive the fill quantity for the larger parent order. But not every metric is simple to construct, so we’ll cover a few more examples here as well.

For each wave, an order arrival (the midpoint price at the time the order arrives) and order end mid (the midpoint price at the time the order ends) are depicted in the image above as well. If the orders are not combined, the arrival performance for each wave will be calculated according to each individual wave’s corresponding arrival price. However, when we stitch the orders together, we will benchmark all executions to the first wave’s arrival price, as shown in Figure 2 below. Similarly, for the “order end mid” benchmark in our TCA, we would benchmark all executions to the final wave’s order end mid.

A more complex consideration is interval VWAP, for which a provider has a number of options in calculating for a stitched order. In the three waves depicted above, which do not overlap, we could choose to benchmark the order to the VWAP over the full trading duration from first wave’s start to final wave’s end. However, it is possible that the waves overlap, which would make this benchmark less useful. In addition, this method doesn’t necessarily represent the VWAP price during the trading interval, as there may be long periods with no trading during the horizon covered by the waves. At BestEx Research, we calculate a stitched order’s interval VWAP performance as the weighted average of individual waves’ interval VWAP performance, weighted by filled notional value. There are other options, however, for evaluating the performance of the waves compared to the VWAP during the order’s duration; clients may prefer to benchmark to the day’s VWAP price as opposed to the interval–if the waves traded over much of the day, for example–or the VWAP from order start to market close, which we provide.

Some traditional measures need to be discarded for stitched orders. For example, a commonly used participation rate metric is the order’s percentage of interval volume. This metric is not useful for determining the participation rate of a stitched order, because the client may not have been trading over the full duration from the first wave’s arrival to the final wave’s end. For this reason, a percentage of interval volume may be heavily skewed downward. Each wave could be trading at 15% of volume, but if they are relatively short compared to the stitched order’s full duration, they’ll represent only a small percentage of interval volume when aggregated.

To calculate order size as a percentage of average daily volume (ADV), however, is straightforward. The total order size as a percentage of ADV would be the total shares filled across all waves divided by the ADV. This metric can look very different for stitched and unstitched orders. If a customer divides an order that is 10% of the product’s ADV into two equal waves, each would measure 5% of ADV in the unstitched order summary while representing the true parent order’s full 10% of ADV in a stitched order summary.  

As illustrated in the above examples, each element of reporting stitched orders must be thoughtfully reengineered to reflect the most useful information as clients review the orders’ characteristics and performance.

Table 1 below shows an excerpt of performance measures for a group of unstitched orders and its corresponding stitched orders. In the table, readers can see that stitching naturally reduces the number of orders and changes the calculation of the order’s size by ADV and realized day volume. In addition, and perhaps most importantly, it changes the view of performance measures like cost vs. arrival price and cost vs. midpoint price at order end. Looking at the cost versus arrival for the original orders, we see an average performance of 6.4 bps, for example, but when we stitch appropriate orders together and assign the first arrival price as the benchmark, we can see the orders may be creating much more impact on average than is detected among unstitched orders. Of course, it could be related to market prices moving away from the trader as well, as we see that the order size is relatively small, but deeper study of the orders would be required to determine whether the orders are creating the price change.

Accessing Stitched Performance in TCA

To create effective views of performance for traders executing in waves, it's essential to have the necessary infrastructure and tools in place. Order stitching is now available in our TCost Summary report in AMS’s Reports tool as an additional tab. Same-day order stitching is currently available for orders traded on the same date, symbol, and side to be stitched, and the additional tab including stitched performance with updated benchmarks will appear as the final tab in this report–all other tabs represent unstitched performance calculations.

There are two “flavors” of stitching available, as illustrated in Figure 3 below. Orders can be stitched across strategies or within strategies, using the Order Stitching drop-down menu to specify. For example, if a client chooses a liquidity seeking strategy for the first wave of an order but frequently changes strategies for later waves and wants all of those waves aggregated as a single order, then stitching across strategies may be appropriate. The trader would choose “Day-Side-Symbol” stitching in this case.

However, if there are multiple orders for the same date, symbol, and side sent to different strategies for different reasons and the trader would like to keep them separate, that’s a possibility as well. Stitching within strategies would be most appropriate in this situation, separating VWAP orders and liquidity-seeking (“Optimal”) orders, for example. The trader would choose “Day-Side-Symbol-Algo” stitching in this case.

While traders may trade larger parent orders in smaller waves over multiple days (rather than just a single trading session as described above), automated multi-day order stitching for performance assessment is not currently available in our TCost Summary report.

Both stitched and unstitched views of performance can be helpful in evaluating the performance of trading strategies. Order stitching provides traders the ability to measure performance for the real parent order they are trading rather than the waves they send to execution algorithms, allowing for better decision-making and optimization. As always, transparency is our aim in providing this information. The additional piece of the puzzle offered by the Order Stitching tab may enhance the ability to evaluate and refine trading strategies effectively.

If you have questions about how to access the Order Stitching functionality or how to interpret statistics provided in one of our Reports, contact your account representative or email us at info@bestexresearch.com.