Dark pools have emerged as a critical part of the execution strategies of buy-side firms large and small. Estimates of their size vary from 9 percent of all U.S. equities trades to nearly 20 percent.
In a bid to boost liquidity, sometime around 2005 dark pools began allowing high-frequency quantitative traders to provide liquidity, which fueled the promise of high volume and quick fills. But high-frequency traders' "synthetic" investment agenda is short-term gain based on market inefficiencies, as opposed to the long-term investment objectives held by their "natural" buy-side counterparts.
Over time, dark pools become "toxic," riddled with gaming risk and missed opportunities for slower-moving institutional orders, many market participants say. One of the most significant risks to trading on dark pools is adverse selection, which occurs when the price of a stock moves against a trader immediately after executing a trade that is part of a larger block. This can lead to implementation shortfall -- the overall block trade is executed at an average price that is unfavorable compared to the optimal price available during the execution time period.
Adverse selection occurs systematically in many dark pools, which is frustrating for many buy-side traders, particularly because the premise of dark pools is that they are safe venues for completing large block trades without exposure to the market. Yet many buy-side traders say they feel compelled to use dark pools because trade size in displayed markets is too small.
"The risks of using dark pools have largely been accepted," says Sang Lee, cofounder and managing partner at Aite Group in Boston. "Using dark pools is a legitimate part of the process. The average trade size in lit markets has more or less disappeared. As an institutional player, if you need to move a large block, your options are fairly limited."
Technology providers and buy-side firms have devised numerous strategies and tools to mitigate these risks. Many dark pool operators and aggregators allow participating firms to deselect interactions with certain order types and counterparties. Other platforms, including ITG's Posit, simply don't accept high-frequency flow. Still, according to Lee, broker dark pools that rely on internalization, which may include principal and quantitative flow, command 70 percent of overall dark pool volume, while agency and block dark pools that are more buy-side focused control just 8 percent and 5 percent, respectively.
To effectively navigate dark pools and achieve best execution, most buy-side firms rely on their brokers for transaction cost analysis (TCA) and for aggregation of dark pools into manageable order tickets, Lee says. But for some firms, broker-provided analytic and execution tools don't provide enough of an edge, so they build their own tools to augment the process, he adds. The scale of a buy-side firm's arsenal for optimizing execution performance, however, varies greatly and can be loosely correlated with its level of assets under management and trading style.
A Hybrid Analysis Style
At Russell Investments in Tacoma, Wash., which manages the Russell 2000 Index and controls $176 billion in assets, the trading technology environment resembles that of a midsize brokerage, according to Jason Lenzo, the firm's head of equity and fixed income. In order to achieve maximum understanding of execution quality, the firm has built a large, real-time tick database and cross-references its in-house risk analysis tools against multiple vendor tools, he reports.
"It really becomes a question of infrastructure," says Lenzo. "Where does liquidity reside? Every day is new, and every second is different. We have the ability to identify all prints by venue, which is very important. We need to be able to assess where executions are happening at all points in time."
Russell built a hybrid infrastructure to analyze trades across all types of execution venues, in as close to real-time as possible, Lenzo continues. He says he uses four risk management packages from vendors, which he declines to identify, to validate against his own internal analysis platform across five major factors: percentage of volume provided or taken from the market, average spread, reaction to volatility, sector tracking bias, and the number of times a print occurs in a day.
"There is nothing we've been able to buy off the shelf that provides the complete solution," Lenzo says. "Regardless of how much quantitative analysis has gone into it, at some point every risk analytics package contains a bias that represents someone's opinion. Somewhere at the core of every system is a belief or philosophy -- someone has made a decision and says, 'If I trade a block of stocks, I'd expect the market to be moved five ticks. Less than five ticks is good, and more than five is bad.' By looking across multiple packages, you can take out some of the philosophical bias that exists in every risk model -- your own included."