An alternative strategy for small, low-urgency orders is to DMA-route limit orders to execution venues and smart routers. The limit orders should be priced at or near the quote and actively managed to control the possibility that the market may move away from the limit price; small-order, spread-capture algorithms do this algorithmically. Day-VWAP algorithms are well-suited for large, low-urgency orders, especially in large-cap and mid-cap stocks.
For high-urgency orders, buy-side traders have fewer execution choices. For small, high-urgency orders in large-cap stocks (corner cube 5 in figure 4) a possible strategy is to send market orders to DMA smart-routers. Large, high-urgency orders in large-cap and mid-cap stocks should consider implementation shortfall algorithms. For large, high-urgency orders in small-cap stocks (corner cube 8), the only option is broker-dealer capital.
The third step in choosing the right execution strategy is to use post-trade execution quality analysis to evaluate the execution choices made and change strategy accordingly. Post-trade execution quality analysis is important both in choosing a particular type of execution strategy, for example day-VWAP over implementation shortfall, and also in comparing the day-VWAP strategies offered by competing trading platforms (not all VWAP engines are the same).
For post-trade execution quality analysis to be useful, a large number of executions must be analysed to ensure the statistical reliability of the estimates, and care must be taken to ensure apples-to-apples comparisons.
Ensuring apples-to-apples comparisons is particularly difficult in this context because the choice of execution strategy depends on the characteristics of the orders: for example, easy orders go to low-touch while difficult orders go to high-touch. In using post-trade execution quality analysis to compare strategies, therefore, one must control for order difficulty.
Trouble with tradition
The traditional way of controlling for order difficulty is to use a trading cost model to derive a cost estimate and then compare the value-added of the actual results to the model estimate. Trading cost models, however, only capture some aspects of order difficulty: for example, order size, average trading volume, stock capitalisation and stock-price volatility. Trading cost models do not capture differences in trading alpha. In comparing execution quality across strategies and platforms, therefore, it is important not only to compare relative to a trading cost model but also relative to an estimate of trading alpha.
Table 1 shows a hypothetical example to illustrate the importance of taking into account trading alpha. In the example, the two strategies have the same value-added relative to the cost model estimates; the orders using strategy 2, however, have higher trading alpha. Strategy 2 works better because it allows the trader to capture some of the higher trading alpha.
Source: Goldman Sachs
1. For buy orders, execution price minus price at order arrival as a % of price at order arrival
2. T-cost model estimate minus actual execution shortfall
3. For buy orders, average closing price over the five trading days following order completion minus price at order arrival, as a % of price at order arrival
George Sofianos is vice-president, Equity Product Strategies, Goldman Sachs & Co., and Peter Sheridan is co-head of Goldman Sachs European Algorithmic Trading. To learn more about choosing the right execution strategy, contact:
peter.sheridan@gs.com,
+44 (0)20 7552 5936
