Liquidnet Preps Four New Algos

The strategies employ a hybrid statistical model that combines historical data about volume-traded and real-time updates. "As we go into the trading day, volume prediction becomes more and more precise," explains Glukhov. "In the beginning [of trading] we have some estimate on what volume will be doing the rest of the day and then, as we get real-time updates about what happens in the market, the model updates," he adds.

The model is self-adjusting and is able to constantly create new trading trajectories for orders based on the incoming real-time market data, from the open liquidity pools, as well as from Liquidnet's own natural buy-side-only matching system and H2O, an aggregator of block liquidity.

The underlying analytical models for the new strategies-VWAP Hybrid, VWAP Blocks, Close and Adaptive Optimum-took years to develop, according to David Fellah, another member of Liquidnet's product management group.

The VWAP Hybrid combines two methods of trading: block and market participation. An order can be executed via the block-trading algorithm to minimize market impact, while maintaining the volume-weighted average price (VWAP) benchmark, Fellah explains.

VWAP Blocks is an implementation shortfall strategy with zero aggressiveness, which will execute an order at the lowest market speed possible following the VWAP benchmark, according to Liquidnet officials. However, if the algorithm finds a block of liquidity to match the order, it will be executed in part or entirely. Through VWAP Blocks, an order can be fully executed during any part of the day. If no matching block is found, then the order will be executed via VWAP, Fellah says.

Adaptive Optimum is a next-generation implementation shortfall strategy that tries to predict the timing and the size of the block execution, designed to reduce impact and timing risk, says Glukhov. It takes into account the open market liquidity and dark block pools, he explains.

The Close algorithm mimics in reverse of the Adaptive Optimum strategy, Fellah says. It is designed to execute orders as block sizes increase toward the close of trading. "So you minimize your risk of missing the close," Fellah says.

The four strategies are in prototype form and in pilot testing with several clients, say officials. Liquidnet hopes the algorithms will lead to more orders crossing internally to minimize impact on customers' orders and to reduce their exposure to risk, Fellah says. To date, nearly 70 percent of orders executed through Adaptive Optimum were crosses internally.

The strategy is currently available in Europe. The Liquidnet development team is working on an Adaptive Optimum algorithm applicable to global portfolio trading.