Algos Take Aim at the Dark

Aggregating Crossing Networks is Really Hard Work

Dark pools are big news these days. Buyside traders want to access and navigate all of them easily. Traders Magazine presents a two-part story about this contentious issue. The first part focuses on efforts to provide traders with efficient access. The second part, which begins on page 66, deals with some of the costs associated with that access.

One step forward and one step back. That’s the situation when it comes to the touchy business of aggregating dark pools. Some brokerages are welcoming rivals into their private realms with open arms. Others are slamming the door in their faces.

No sooner had Lehman Brothers and Fidelity Brokerage Company announced an agreement to link their respective crossing systems than Pipeline Trading Systems, operator of one of the major “independent” dark pools, revoked access to the broker-sponsored algorithms used by buyside firms.

The moves by these firms highlight the stop-and-go nature of the industry-wide effort to ease access to the ever-growing number of dark pools.

“Access is the name of the game right now,” says Michael Bleich, head of liquidity strategy for U.S. equities at Lehman Brothers. “Fragmentation is the problem brokers need to solve on behalf of their clients.”

Most of today’s 30-plus crossing systems didn’t even exist a year ago. They mushroomed in response to traders’ needs for more efficient methods of trading large blocks of stock.

Dark Algos

Traders got what they wanted, but are now reeling under too many choices. They must take into account all of these different systems when seeking liquidity. They would like the process of accessing dark liquidity to be easier and swifter.

Access can take many forms. Brokers have built algorithms to access multiple dark pools at once. Many of these “dark algos” use smart order routing to intelligently extract liquidity from those venues. Dark pools can also link to one another directly, as Lehman and Fidelity did. Finally, brokers can send order flow to utility-like consortia such as LeveL and BIDS. These platforms are designed to pool diverse forms of liquidity for customers through sponsoring brokers.

Today, algorithms are the most popular option, in part because buyside firms are comfortable trading in the displayed market via algos. Close to two-dozen broker-dealers have developed algorithms that search multiple dark pools, and new algos are launching every few weeks. Six months ago, only a handful of dark algos existed.

The New Race

Algorithms that hunt for non-displayed liquidity are also popu-

lar because brokers can expect a windfall of business as more buysiders turn to algos to access dark liquidity.

“Clients will eventually use algos almost exclusively to access these pools of liquidity,” says Tim Reilly, head of U.S. electronic execution sales at Citigroup. “It’s time-consuming and complicated to go from one pool to another.”

UBS, Lehman and Banc of America Securities all launched dark-liquidity algos last fall. All say these algos were the quickest to gain traction and attract order flow among their various algorithmic offerings. BofA’s Ambush algo is the firm’s “most successful algorithm to date,” says Bill Harts, head of strategy for equities at Banc of America Securities. UBS’s Tap algo already accounts for more than 30 percent of UBS’s algorithmic executions, says Will Sterling, that firm’s global head of electronic equities trading.

Brokers can also charge more for executions through dark algos than they can for algorithms that access the displayed markets. Brokers’ fees for customers using their dark algos range from 1 cent per share to a high of around 2.5 cents per share, according to industry participants. That’s significantly higher than algos that piece orders out to the public markets.

The main problem is that most of the bulge-bracket pools are closed to all non-proprietary algos, making access a hassle for traders. Goldman Sachs, Morgan Stanley, Merrill Lynch, UBS and others point out that aggregating undisplayed liquidity is crucial for buysiders, but insist their pools remain proprietary-at least for the time being.

At JPMorgan, which doesn’t operate a dark pool but offers several dark algos and direct market access to more than 15 pools via its Neovest front-end, the frustration is palpable.

“We’ve got to get together [as an industry] and open ourselves up and provide access, so that the algorithms can start to play in all these different dark pools of liquidity,” David Conner, head of electronic client solutions sales at JPMorgan, told a recent gathering of buyside traders. “Fragmentation has arguably gotten worse.”

Tracking Dark Pools

That fragmentation makes the pursuit of liquidity more complex for the buyside. Buyside traders need to track the rule sets and requirements of different dark pools, as well as the types of liquidity they attract. Traders must figure out when to enter and exit various venues and how long to keep orders within certain pools.

“The market is saturated with dark pools, and some are still in the development stage,” says Bob Gauvain, director of U.S. trading at Pioneer Global Asset Management. “You must search different pools, but then you have to wonder: Am I picking the right or wrong broker? Who has the flow?”

“If a 1-million-share print goes up and you’re not in Broker X because you’re in Broker Y, that affects your executions and how you’re evaluated,” Gauvain adds.

Pioneer, which oversees $65 billion in U.S. equities, has dark-liquidity algorithms from most of the major brokers embedded in its order management system. Pioneer’s traders access dark pools directly, by submitting orders into the crossing systems, and also use several dark algos to ease the work flow and tap into multiple destinations.

The problem of access is not trivial. Trading in dark pools is growing smartly and is expected to continue. Since 2005, volume executed on crossing systems has risen 50 percent, according to research firm TABB Group. Today, about 512 million shares per day cross within dark pools.

Algo Aggregation

TABB expects a daily average of 1.5 billion shares to change hands through crossing systems and internal broker-dealer dark pools in 2010. That’s likely to represent 15 percent of the overall market in 2010, almost a 50 percent increase over this year’s 9.4 percent.

With so much volume at stake, brokers are primed to offer aggregated, efficient access to that liquidity. Broker algorithms usually access three to eight crossing systems, including the firm’s own dark pool if one exists. They also route flow via hidden and reserve order types to NYSE Arca, BATS and Nasdaq to take advantage of dark liquidity in the displayed markets.

All broker algorithms route to NYFIX Millennium. Many of the largest provide smart-order-routed liquidity on Liquidnet H2O. A smaller number-around 12-accessed Pipeline until last month. As broker consortia and exchanges such as Nasdaq launch their own crossing products, those are also being added to brokers’ algorithmic itineraries.

More Access

The dark pools clamped shut to other brokers’ algos are those run by the big brokers. However, many industry participants expect that to change. Just as front-ends such as Goldman’s REDIPlus and ITG’s Triton are now providing access to algorithms from other brokers, so too, the argument goes, will brokers’ dark-liquidity pools become accessible to customers routing through competing firms. The reason: Customers will demand it.

Indeed, the Lehman-Fidelity linkage deal followed on the heels of a similar arrangement between Instinet and Credit Suisse. Last fall the two brokers agreed that each firm’s dark algorithms could have reciprocal access to their own dark pool-Continuous Block Crossing, in the case of Instinet, and Credit Suisse’s CrossFinder. Instinet also allows several other brokers to access CBX.

“We need to access every pool that provides sufficient liquidity,” says David Brooks, director of equity trading at Boston Company Asset Management, which has $70 billion in equities globally. “Not accessing every pool could hamper our ability to execute well.” He and his traders currently place orders directly into dark pools and use several broker algorithms.

Brooks adds that his desk will eventually be obliged to use algos much more to simultaneously access non-displayed liquidity. “Then we’ll adjust those algorithmic strategies accordingly, based on fill rates, the speed of execution and other execution factors,” he says. BCAM crosses just under 10 percent of its U.S. volume, but expects that to double in the next couple of years.

Headline Numbers

The issue of access is not just one of quantity, but also of quality. Simply adding destinations to a dark algo’s roster isn’t enough-and it may not result in the best executions for customers. Smart order routing is considered increasingly vital.

“The headline number can be very misleading,” Lehman’s Bleich says. “What you’re really interested in is getting a trade done, not the number of pools accessed. You could have access to a large number of pools, but if many of them have little liquidity, they’re not adding much value.”

Even if access to more destinations increases an algo’s fill rate, that may not benefit institutions with block trades. An algorithm that isn’t careful about which pools it accesses and how it places orders could signal the orders it’s working to those trying to game dark pools.

“Some people say you can’t see the quotes [in dark pools], so there’s no market impact,” says UBS’s Sterling. “That’s a very dangerous connection. There are many firms trading in parallel in dark pools and public markets, so if I take a bunch of liquidity out of a dark book, it would be naive of me to think that would have no effect on how people are trading in the rest of the market.”

Dark algos must do more than simply represent orders in dark pools, notes Carl Carrie, head of product development in the electronic client services group of JPMorgan. “The new generation of these algos look at liquidity more continuously in how it migrates from the displayed market into the dark and back and forth,” Carrie says. “These algos factor in the implicit cost of trading and can dynamically make adjustments for gaming in displayed and dark pools.”

Competition between dark algos also creates stark choices for some brokers. Those with dark algos and successful dark pools, such as ITG and NYFIX, could face a potential dilemma.

By accessing other dark pools through its DarkServer algo, ITG, for instance, adds to the liquidity in its competitors’ venues. “It’s something you struggle with when you have an aggregator and some form of an ATS: Are you beginning to validate other ATSs by routing order flow there?” says Tony Huck, head of algorithmic trading at ITG.


For ITG, Nasdaq’s decision to launch a scheduled intraday crossing product threw this conflict into high relief. Nasdaq’s planned crosses are designed-and pitched-to the exchange’s sellside members and their customers as a direct, low-cost competitor to ITG’s POSIT Match.

But despite the competition, ITG says aggregating liquidity for clients is its key concern. ITG will connect to Nasdaq’s Intraday Cross “if customers ask for access and there is liquidity in the system,” Huck says.

However, conflicts involving access and algorithms are far from over. Pipeline’s recent decision highlights another-and potentially more critical-source of discord between dark pools and the algos knocking on their doors.

Pipeline last month decided to prevent buyside firms from accessing Pipeline through broker-sponsored algorithms because those algos, it said, did not necessarily provide customers with the best executions. Pipeline’s concern was that brokers have financial incentives to divert their algos away from Pipeline, which is an expensive venue for executions, and toward cheaper pastures.

Access Fees

With third-party broker algorithms, the broker pays the fee for executions, not the customer,” says Fred Federspiel, Pipeline’s founder. “So there’s a conflict present. The broker has all of the pain and none of the gain.” In contrast, customers who access Pipeline directly pay the fee for executions themselves, and therefore control the search for liquidity.

Pipeline has confronted this pricing issue for months. Last fall, ITG was reported to have shut off access to Pipeline for some customers. Federspiel concedes that “it looks like that happened.” He says another major broker terminated access to Pipeline for three weeks in the winter and then began sending flow again, but was down 90 percent from its earlier level.

Michael Plunkett, president for North America of agency broker Instinet, suggests that dark-pool access fees may be an area for the Securities and Exchange Commission to address, as the regulator did in the displayed market. “We’re starting to see a replay of the whole issue with ECNs and best execution and access fees, but in the dark space,” he says.

Solving this problem for dark pools could create a more even playing field by eliminating discrepancies in the cost of access, Plunkett says. That could ultimately open up access to dark venues and enable customers to compare the execution quality of brokers’ dark algos more fairly.

At least one broker is optimistic that a market-based solution to the larger problem of fragmentation will ultimately be found. “One day all the sellside dark pools will be consolidated,” Andrew Silverman, head of U.S. electronic trading distribution at Morgan Stanley, told a gathering of buyside traders. “The question is, when? Until then, there will a lot of pain.”


Wooing the Algos

Crossing systems are intensifying their efforts to get dark algorithms to route orders to them. Those algos offer a way for dark pools to attract more flow and improve their fill rates. They could also become the main source of flow as buyside customers shift from accessing crossing systems directly to using algorithms that provide smart routing to multiple dark pools. LeveL, a consortium owned by five large brokers, gets the vast majority of its order flow from four dark algorithms, according to Whit Conary, LeveL’s president. He expects more algos to connect to the system in the next quarter. BIDS, a consortium whose forthcoming crossing system is pitched toward block trades, also expects algorithms to supply a lot of its flow. “We invite all comers, whether you’re an algo or broker, an owner or non-owner,” says chief executive Tim Mahoney. ConvergEx Cross, the crossing system within BNY ConvergEx Group, is open to all algos that post liquidity on its system. “We’re talking to all algo providers, and we hope that by end of the year they’ll all look to us for liquidity,” says Gary Ardell, head of the financial engineering and advanced trading solutions unit at BNY ConvergEx Group. To increase executions, some dark pools send indications or messages to brokers’ algorithmic engines that their systems have resident liquidity in particular names. That alerts algos hunting for liquidity in that name to submit an order into the pool, where it may match up against an order that is already there. This electronic messaging goes directly to the engine running the algorithm, without human intervention. These indications serve to bring together naturals in an automated fashion-with less latency and before blocks get chopped into smaller pieces as they search for liquidity. “The ability of dark pools to reach out and ping algorithms to drive routing decisions is extremely valuable functionality,” says David Brooks, director of equity trading at Boston Company Asset Management. “It will increase their effectiveness.” NYFIX Millennium, Liquidnet H2O and ITG’s Block Alert send out alerts. Pipeline did so until recently. BIDS is considering this option; LeveL and ConvergEx Cross are not. -N.M.


ITG’s Flip-Flop

Before Pipeline abruptly decided buysiders couldn’t use broker-provided algos to access its crossing system, there was ITG. Last fall ITG told Credit Suisse that the Guerrilla dark algorithm it provided customers was no longer welcome in POSIT, ITG’s headline crossing system. Other brokers whose customers used their dark algos to access POSIT also had to cease and desist. ITG has since retreated from that stance-quietly. Brokers can now give customers access to POSIT via their algorithms, although they must disclose certain information to ITG, such as the approximate percentage of order flow they are making available to POSIT through that algo. Manny Santayana, head of sales in Credit Suisse’s Advanced Execution Services group, attributes the initial decision to competition. He says ITG didn’t want buyside traders accessing POSIT through algorithms other than its own DarkServer algo. “We’re dominant in that space-and to make a long story short, they were trying to catch up with us,” Santayana says. Credit Suisse, one of the most popular algo providers to the buyside, launched Guerrilla in 2004. Tony Huck, head of ITG’s algorithmic trading, counters that “POSIT has valuable brand recognition and unique liquidity. ITG’s main priority is ensuring that customers have meaningful access to our liquidity, whether they come through a broker algorithm or directly to POSIT.” Industry observers speculate that ITG also didn’t want customers heading to a broker that could provide cheaper access to POSIT than they could get directly from ITG. “Cost arbitrage was an issue,” says an exec at a large broker-dealer. POSIT remained open to buyside firms that accessed it through ITG, and to about 100 brokers. Brokers could submit orders into POSIT through their cash, program trading and proprietary trading desks. They just couldn’t give customers direct desktop access to POSIT via their algorithms. But now, a handful of brokers are offering clients dark algos on their desktops with POSIT access. They apparently pay ITG 1.5 cents per share, a higher fee than the brokers pay through their sales desks. Several big brokers also have different versions of dark algos for clients-one DMA version with no POSIT access and another version, available only through the broker’s sales desk, with POSIT access. Brokers may charge different fees for those algos. -N.M.