The following story originally appeared in the August 20013 edition of Traders Magazine
Cowen and Company is making tools available to its buyside customers that it believes will help them when deciding which trading algorithm to use.
The new service marries technology that predicts the price of a given stock in the very short term with technology that switches between several Cowen algorithms. Based on the trader’s level of aggressiveness and the expected price swing in the stock, the switching technology picks the optimum trading algorithm.
The net result is Best Execution Switching Technology or BEST, an algorithm that reacts and adjusts in real time on behalf of a buyside trader’s desired trading strategy, said John Cosenza, co-head of electronic trading at Cowen/ATM. Algorithmic Trading Management, or ATM, is an electronic trading firm Cowen purchased in 2012. This dynamic and automatic functionality replaces the manual process of changing algorithms or trading strategies during the life of a buyside order, Cosenza added.
“BEST adjusts itself dynamically and logically as a stock is traded,” he told Traders Magazine. “The dynamic functionality is essentially an evolution of what we’ve seen buyside traders do manually with different algo suites.”
The algorithm has two core components: first, a proprietary short-term alpha model that generates a prediction about whether the stock will go up or down and a measure of confidence of that forecast; and second, switching-engine capability that dynamically changes algorithmic trading strategies such as VWAP, implementation shortfall, etc.
The switching is based upon user inputs for level of aggression, stock price movement, the BEST score and changing market conditions.
Cosenza explained that the algorithm operates by first computing the BEST score from the short-term alpha model for a stock. The alpha model looks at technical and relative value factors in determining a BEST Score, which ranges from -100 for a stock expected to fall in price to 100 for a stock expected to rise in price. As an example, relative value can be defined as the stock price movement versus sector ETF price movement. Then the algorithm processes this score as a factor into the decision-making process, depending on the strength of the score.
The data is presented to the trader, who can then input a number indicating how aggressively the algorithm should be searching for the stock, on a scale of one (least aggressive) to five (most aggressive.) The user-defined “aggression” parameter determines how much of a factor the BEST score plays into the switching engine’s decision to switch from one particular strategy to another strategy.
“The BEST algorithm was programmed using information we gathered through dialogue with execution traders and extensive analysis of FIX traffic logs that illustrate customer behavior,” Cosenza said.
Once the alpha and BEST score are computed, the parent order is divvied into child orders and routed out via the execution-switching logic. The logic then constantly monitors and alters just which algorithmic strategy works best, given the buyside traders’ inputs, such as aggression level, market conditions and stock price. Cosenza explained that an order that might first be handled via percentage of volume could then be automatically switched to implementation shortfall or liquidity seeking strategies, and so on.
For example, suppose a trader selects an aggression level of five on a buy order of 100,000 shares of IBM when the stock price upon entry of the order is 192.60. If the stock drops to 192.50 (in the money) and the BEST score in real time reads 90 (a strong signal the stock is going up), BEST will use a liquidity-seeking strategy to very aggressively to take advantage of current prices.
In the same scenario, had the BEST score in real time read -90 (strong signal the stock is going down), BEST would use a participation algorithm between 10 percent and 20 percent, allowing the stock price to further drop.
See Chart:How it Works
At least one buysider likes the idea of the BEST system. William Lishman, head of U.S. trading at Schroders Investment in New York, told Traders Magazine that any extra information regarding how a stock may move is useful. He uses the BEST algorithm.
“Predictive analytics focuses on the very short term and can be a very useful tool when coupled with a longer-term trading strategy,” Lishman said.
He added that when traders use algos, it is important not to be too predictable, for obvious reasons, and having an algo that uses switching technology and that has predictive analytics as one of its many inputs can help save a trader from having to make as many manual changes as they otherwise would.
“The trader still needs to control the order and will still monitor and make changes as necessary, but a good switching strategy can help take some of the pressure off by making changes based on very short-term market dynamics,” he said. “With big changes in market structure and the trading environment over the last few years, we are constantly looking for any advantages we can get, and are actively working with Cowen to further customize this technology to enhance our trading capabilities.”
