Portfolio Trading – Plotting Your Optimization
Big index funds and mutual funds that trade passively are popular these days, but moving large positions without incurring market impact can be challenging.
Pension funds and index funds that periodically rebalance their portfolios once a month or once a quarter can face similar portfolio trading risks.
FlexPTS (Portfolio Trade Scheduler)
To solve this problem, FlexTrade built FlexPTS, a sophisticated optimization tool that determines the best trading schedule for portfolios. After it’s given a portfolio trading list with targets to buy and to sell, FlexPTS generates a schedule of buy and sell orders divided into 15-minute windows, such that the target sizes are met at the end of the trade. Representing the third leg of the company’s FlexEdge advanced analytics offering, FlexPTS is designed for multi-day time windows and managing global portfolios. IBM’s renowned Mathematical Optimization Library ILOG CPLEX is used under the hood.
The product is targeted at buy- or sell-side firms that trade over a half-day, single day or multi-day period of time because parts of the trade list are illiquid.
“Illiquid means that the amount that needs to be traded in the names of the portfolio is high relative to the average daily volumes,” explains Ran Hilai, vice president of portfolio optimization at FlexTrade.
How FlexPTS Works
To avoid market impact costs, buy and sell-side firms can run their ideas through FlexPTS’s optimizer, which then generates a trading schedule of how much in each name to buy and sell. Portfolio-based trade scheduling is provided through state-of-the-art optimization algorithms, portfolio risk analysis and market impact modeling.
Utilized by program trading desks, FlexPTS employs an implementation shortfall portfolio trading algorithm to help traders attain the arrival price benchmark.
Implementation shortfall is a measure of the difference between the average execution price and the price when the order arrives on the trading desk.
In addition, traders need to minimize price risk, but these can be conflicting goals, explains Hilai. “If someone wants to minimize the cost relative to arrival price, they need to take a long time by trading passively. But if someone wants to trade close to the arrival price in order to reduce their risk, they need to trade aggressively before the price changes,” he adds. “Trading slowly results in a low market impact but incurs a high price risk. However, trading quickly reduces the price risk, but results in a high impact,” he explains.
The goal of FlexPTS is to find the schedule representing the best trade- off between these two conflicting goals, while keeping the client’s constraints. For example, clients can set their own constraints, such as,
- Cash constraint: a limitation set on the imbalance between the buy and sell values of the yet unexecuted part of the portfolio.
- Participation constraint: a trade limit in terms of a percent of the window average daily volume (ADV) applied to each 15 minute trade interval.
Portfolio Risk Control
While implementation shortfall can be applied to single-name trading, scheduling an entire portfolio is a more effective way to reduce risk, especially if it’s a long/short portfolio. This is true because the scheduler uses observed price correlations to reduce the risk of a trade. “If this is a fund rebalance, you typically have a buy and sell side portfolio,” illustrates Hilai. “By sticking to a trading schedule, you can get into the portfolio and do some trading early that can improve the tracking error for the buys relative to the sells,” says Hilai. “Then, once the tracking error is reduced, you can take more time and trade more passively since you have hedged into the portfolio and don’t need to spend more time on hedging,” explains Hilai. “Risk control is the whole point.”
FlexPTS enables the trader to control risk across the entire portfolio. A trade scheduler estimates the volatility of any portfolio using a portfolio risk model, which, in turn, provides estimates of the volatility of each stock and the correlation between risk factors, such as sectors and industries, fundamentals and statistical factors.
FlexPTS incorporates the daily risk model from Northfield Information Services, Inc., or this model can be replaced by any client’s or third-party’s risk model. The risk model is looking at the whole portfolio and every name has correlations with other names in the portfolio. “Some names are removed immediately, because doing that will remove the risk.
Scheduling Your Trades
In terms of portfolio scheduling, if a trader gets rid of the names that are adding risk, then returns will improve. However, returns, especially over the long term, depend on the portfolio manager’s selection of holdings. FlexPTS merely deals with the trading cost relative to the arrival price. Risk has a lot to do with the correlation between names in the portfolio,” explains Hilai.
As a result of running FlexPTS, the trader will get a trading schedule for completing the trade. While PTS does not select the names to trade, it merely suggests a trade. “If the trader goes line-by-line, the names that will be finished early will be the liquid names, while the illiquid names will be finished last, explains Hilai.
If traders deviate from the suggested schedule due to reasons such as a dark-pool match in an illiquid name, the FlexPTS optimizer can be invoked multiple times to re-optimize the remaining trade schedule.
As opposed to most available algorithms, FlexPTS, which handles part-day, single day or multi-day schedules, can automatically determine the optimal length of the trade window.
Under the Hood
Since FlexPTS is an optimizer, it has a utility function with constraints. The utility function is defined as the minimization of the expected market impact cost plus the expected market risk. The utility is built with two functions:
- The expected market impact based on the model that FlexPTS is using over the whole portfolio and duration of the trade.
- FlexPTS puts the risk, which is the square root of variance in the utility function and trades it off against the expected cost. “Other systems may use the variance because it’s mathematically easier but gives inferior results,” according to Hilai.
The Process in Action
When measuring risk across the entire portfolio, a given one-day schedule could instruct the trader to execute 26 different trades at 15-minute intervals. The risk model will provide the variance. Then FlexPTS would add the variances of all 26 trades, and then take the square root of the sum. This results in the expected variance relative to arrival price, according to Hilai.
At the end of the schedule, whether it’s the buy-side or sell-side, the trader has to complete the total execution quantity required by the portfolio manager, says Hilai. “Those are the constraints and the utility function is there to minimize the Value at Risk,” he says.
To preserve the confidentiality of their strategy, buy side traders can optimize their trades without exposing their entire list to the sell side. Rather than send their entire list, trades can be forwarded to sell-side desks or algorithmic servers in smaller orders following the optimal schedule provided by FlexPTS.
Users can access FlexPTS actions from the toolbar in FlexTRADER EMS. After a portfolio is optimized, the resulting schedule can become an integrated part of any execution strategy in FlexTRADER, including the use of custom trading algorithms and all available dark pools. During trading, various FlexPTS analytics constantly update on the FlexPTS front-end.
For more information about FlexPTS and its integration with your system, please contact us at email@example.com.