Trading models seems to be the new fashion on Wall Street. Institutional investors, hedge funds, vendors and brokers all are developing trading models to carve large orders into smaller pieces and execute them in accordance with specific trading strategies.
But how do firms select their models? And how do they know which model works best for which security or during which market conditions? Well ...
According to Tabb Group research, 57 percent of institutional investors are experimenting with algorithms and 48 percent of hedge funds select their broker-sponsored trading algorithms based upon their broker relationship, rather than on how well the algorithms perform. If we look at the vast amount of order flow being transacted via algorithms - approximately 7 percent of institutional investor order flow and 10 percent of hedge fund order flow - it is obvious that we need a better way to measure how these models perform and determine which models to employ. The solution, however, is easier stated than implemented.
The traditional way of measuring broker execution quality is transaction cost analysis (TCA). These analytics compare a firm's trading history to the universe of executed transactions and, through various proprietary technology and benchmarks, ascribe execution-quality metrics to various brokers.
The problem with many TCA technologies is that the models were designed for another age, when the buy-side trader had a limited set of trading options; the majority of buy-side trading was delegated to the broker; and market orders were not shredded into retail-size parcels and executed electronically.
But where does that leave the trader in today's increasingly model-driven world? To develop effective TCA for the trader, much must be changed.
First, trading TCA must be implemented across the organization. At the outset, all trades are not alike and all orders cannot be analyzed using the same criterion. Some orders are time-sensitive, some are cost-sensitive and some executions must be linked to the open, close or average price. This information must be conveyed electronically from the portfolio manager to the trader to the market. Connectivity must be developed to communicate trading instructions from the portfolio management system to the order management system to the TCA analytics - and back. This way, the tool understands the order and can analyze it consistently against an appropriate benchmark.
Timeliness is the next challenge. A trader cannot remember the market conditions in which a trade was executed four months ago. To support the trader, feedback needs to occur in real time or daily - at the very least, weekly. Preferably, TCA technology would monitor in-flight orders and notify the trader when an execution goes awry so the trader could change tactics before the order is filled, rather than justify an inappropriate execution to an irate portfolio manager. To accomplish this, however, buy-side traders not only need the monitoring capabilities; they also need real-time executions and market data and more sophisticated TCA tools.
Additionally, as TCA technology becomes more advanced, it must understand the algorithms and strategies, and track the deployment conditions. This will enable traders to better understand models' strengths and weaknesses, and how to employ them and match them with the appropriate stock and trading conditions.
As model-based trading continues to expand and the buy side takes more control over its trading, TCA will become increasingly important. The new TCA will be a full-fledged tool on the trading desk that helps traders select the appropriate strategy for the job.
While models may be the new Wall Street fashion, TCA will be the new "black" - the universal color (product) that models will be measured against and that no trading desk will be without.
Larry Tabb is founder and CEO of Westborough, Mass.-based The Tabb Group, a financial-markets strategic-advisory firm. firstname.lastname@example.org