ITG, an independent execution and research broker, includes software quality control as part of its product development for its algorithms, including its new Dynamic Close algorithm, which empowers traders to tap the sizeable liquidity available in the NYSE and Nasdaq closing auctions in a more precise manner than competing algorithms.

“All of our algorithms, including the Dynamic Close, begin with extensive research and development,” said Jeff Bacidore, ITG’s managing director and head of algorithmic trading. “In the case of the Dynamic Close, we did an exhaustive study of the market dynamics around the close of trading—including both open market trading and the closing auction itself.”

ITG’s Dynamic Close algorithm recognizes that the majority of auction impact occurs at the imbalance announcement and it intelligently pursues open market liquidity in the pre-cutoff period.

It’s the first closing algorithm that provides traders with both a flow and a rebalance setting, recognizing the tradeoff between minimizing slippage to the close in the former and reducing implementation shortfall in the latter, according to ITG.

“We consider nuances of the markets such as what data do the exchanges make available, what types of orders are allowed and how that differs by exchange,” said Bacidore. “In doing the analysis for the Dynamic Close algorithm, we learned that the approach taken by traditional close algorithms was actually incorrect, and a completely new approach was needed.”

Once it has completed the research phase, ITG uses the results of the research to inform the design and implementation of the algorithm.

“After the algorithm is implemented, it goes through extensive testing before eventually being deployed to clients,” said Bacidore. “In many cases, including with the Dynamic Close algorithm, we publish the results of our research externally to give clients transparency into why the algo behaves the way it does.”

ITG’s Dynamic Close algorithm allows traders to choose between two different modes: a rebalance mode aimed toward minimizing implementation shortfall, and a flow mode that attempts to outperform the closing price benchmark.

“Our algorithms are carefully calibrated via empirical analysis and user feedback to trade effectively out of the box,” Bacidore said. “But virtually all of our algorithms allow clients to set parameters to more closely align the algorithm’s behavior to the client’s specific strategy on a given order.”

The Financial Industry Regulatory Authority (Finra), a self-regulatory organization of the securities industry, in its 2013 priorities letter to members, said that it “continues to be concerned about how firms are supervising the development of algorithms and trading systems”.

Finra said it “will continue to assess whether firms have adequate testing and control related to high-frequency trading and other algorithmic trading strategies”.

Algorithmic trading refers to the practice of designing an algorithm that is capable of executing pre-programmed trading instructions automatically and without human involvement.

“The move towards algorithmic trading in the securities industry continues to be an area of concern for Finra,” said Richard Levin, a counsel at law firm BakerHostetler, in a client memo. “The ease of use and market risk factors associated with algorithmic trading make this practice appealing to market participants. However, automated trading raises significant concerns due to the perceived lack of a human fail-safe.”

Potential areas of review by Finra will include whether firms conduct separate, independent and robust pre-implementation testing of algorithms and trading systems, and whether the firm controls changes made after an algorithm and trading system is placed into production.

In addition, Finra will focus on whether broker-dealers have kill switches in place.