Commentary

Elaine Wah

Modern Markets, Modern Metrics - A Blog By IEX

In this blog by IEX's Elaine Wah, the newest public exchange looks to refute public claims that the metrics it uses are designed to inflate its own volume numbers and mislead people.

Traders Poll

Do you think it's a good idea to conduct an access fee pilot to assess the pricing models used by many trading venues?

Yes

67%

No

0%

Should have had a pilot program a long time ago.

33%

Free Site Registration

August 14, 2007

Algo Crossing Gains Sophistication

By Nina Mehta

Crossing engines devoted to algorithmic flow are becoming more popular. Merrill Lynch and Morgan Stanley, the only two firms to deploy these niche dark pools intraday, report greater usage and growing volumes. The newfound popularity of the systems comes as more flow is executed via trading algorithms and customers seek to avoid the costs of going to the public market.

"The majority of our customers are now accessing AXP," said Jarrod Yuster, global head of electronic trading at Merrill Lynch. Merrill's AXP, or Algorithmic Crossing Product, was the giant broker's first ATS, launched in January 2006 within the firm's algorithmic trading platform.

AXP crosses algorithmic orders from Merrill's extensive customer base as well as the firm's internal trading desks. Customers got access to the crossing system last October.

Crossing systems geared specifically to algos enable orders to be crossed for overlapping time periods, as opposed to dark pools that match discrete orders at a price within the national best bid or offer.

Morgan Stanley's ATS1 for algo orders was launched back in 1999. It crosses overlapping portions of algorithmic orders with the same or different benchmarks over small periods of time. The execution price is the volume-weighted average price for that brief period.

According to Andrew Silverman, head of Morgan Stanley's U.S. electronic trading distribution, the firm expects to institute changes to ATS1's matching engine this fall that will increase the system's match rate. Toward the end of the year, the broker also plans to allow customers to rest orders in ATS1 to match up against algorithmic orders in the system.

Merrill's AXP crosses pieces of algo orders at execution prices that are not necessarily the VWAP for that interval. For example, if one customer has a 100,000-share sell order in IBM at the VWAP and another is buying 200,000 shares of IBM at an arrival price benchmark, AXP's matching engine will determine whether and when to cross small pieces of those orders and at what price that satisfies both customers.

The AXP engine has two chief benefits, according to Yuster. The first is that crossing orders at the "submission level" (as orders are released into the marketplace), rather than the larger "parent level," enables clients to cancel orders during the day. The other is that crossing orders with different benchmarks over very small periods of time, such as two or three minutes, increases the platform's fill rate.

The engine takes into account the limit prices and constraints on the two orders, the stock's trading level, volatility and other factors in determining the execution price and crossing frequency, Yuster said.

Instinet's VWAP cross is the industry's third algo crossing engine. The dark pool is a pair of pre-open crosses that execute full-day matching orders at the day's VWAP. If one customer, for example, wants to buy 250,000 shares of IBM and another wants to sell 150,000 shares of IBM, the engine would cross 150,000 shares at the full-day VWAP. Instinet says it executes 18 million shares per day in its cross. Morgan Stanley and Merrill Lynch declined to disclose their volumes.