While high frequency traders may be a step ahead in the quest for ultra-short-term alpha, institutional traders possess a significant advantage in trade execution based on their knowledge of order size and origin. The battle for supremacy in low latency market access is unlikely to provide a good return on investment for the institutional desk, but fighting this battle is unnecessary, as big data techniques can be deployed to control algorithm parameters and minimize the effects of HFT on institutional trade execution. As portfolio managers seek to capitalize on long-term alpha opportunities, the successful leaders will be the ones able to maximize their competitive edge by focusing on exploiting their natural information advantage to schedule trade execution intelligently.
The Problem of Adverse Selection
Most high frequency trading strategies exploit the option value of small orders that are placed on the market when institutional orders are broken down for algorithmic execution – the option to fill or not fill each subsequent order has value and can be monetized using a short-term alpha model. This principle results in algorithmic trades executing more slowly before and during adverse price trends, and too quickly for easier trades executed before further price improvements have been captured. This is the adverse selection problem.
The cost of adverse selection can far exceed the profits on the high-frequency trades that were executed because the delay on an institutional trade’s execution is not limited to the short-term trading horizon of the HFT trader. In some cases, if price moves adversely, a trade may need to be canceled resulting in opportunity costs. Since trading is a zero sum game, when an institution incurs losses from adverse selection far larger than the HFT’s profit, the benefit goes to a longer-term liquidity provider who is willing to be patient and hold inventory over a period of time comparable to the execution horizon. This can be a hedge fund or cash desk, but it can also be another institution that is willing to be patient and act as liquidity provider. Understanding when a portfolio manager can benefit from patient trade execution is perhaps the greatest potential benefit of deploying quantitative optimization methods in trade execution.
Empower the Trader
To take control of these challenges and identify opportunities for optimization, some institutional desks aim to develop their own trading platforms and use low-latency market access to gain time priority over their competitors, if not over the high frequency trading firms themselves. This approach is expensive and presents additional risks. The very investments made by the buy-side to try to compete with high frequency traders can have unintended consequences that can in some cases increase vulnerability to high frequency trading. For example, algorithms that react to quote changes several layers deep in the order book can fall prey to high frequency trading manipulation strategies.
As an alternative approach to give portfolio managers the upper hand, CIOs can play a pivotal role in preparing their firms for competition with high frequency trading strategies and the challenges of adverse selection by empowering Head Traders to make technology decisions that optimize the execution process with attention to the portfolio manager’s risk and cash management needs. Optimization requires three basic elements:
1) Define an execution objective that is directly related to portfolio performance.
2) Translate the portfolio managers' risk aversion and cash needs into explicit constraints.
3) Adopt a methodology for performance measurement and attribution.
By focusing on the optimization of trade execution and empowering the trader, institutional asset managers can pursue alpha by exploiting two data sources that give them a natural advantage over their non-institutional counterparties. First, the real time results of algorithmic execution as the trade is being executed can reveal supply and demand imbalances when compared with a model built from historical data from the same algorithm; this information can be monetized through predictive algorithm control. Second, given a historical dataset of trade origination events, it is possible to build a model to predict trade urgency and recommend an optimal execution strategy for new trades, a technique we have termed "alpha profiling.”
What is the ROI on big data efforts in execution optimization?
Portfolio managers’ primary focus is always going to be to get the right stocks in the portfolio – a stock’s performance versus the benchmark over its portfolio lifespan is far larger than the execution cost. However, execution costs end up making a big difference in a fund’s ranking among peers. For example, an achievable 70bp gain over 5 years is enough to improve the ranking of a fund in the large cap value category from the 60th to the 75th percentile, a significant jump. By working with CIOs, buy-side traders can take advantage of the data already available to them, surmount the adverse selection challenge, and optimize fund performance in the new market ecosystem.
Henri Waelbroeck, Ph.D., serves as Global Head of Research at Portware. Previously he was Director of Research for Pipeline Financial Group, co-founded Adaptive Technologies Inc, and served as Research Professor at the Institute for Nuclear Sciences at UNAM, Mexico. He leads Portware's Alpha Vision research applying machine learning to optimize execution management. Mr. Waelbroeck holds an Engineering Physics degree from the Free University of Brussels and a PhD in Physics from the University of Texas Austin.