In late 2008, Ben Sylvester stumbled upon an idea that would change the way many investment decisions are put to work at JPMorgan Asset Management. His insight was that taking into account a portfolio manager’s trading history, rather than focusing solely on a stock’s quote, could improve performance.
The idea began when Sylvester was spending part of his "garden leave" before joining JPMorgan Asset Management watching the World Series of Poker and other TV shows. When he came across a program on Navy SEAL snipers,
he settled in for what he thought would be "blood, gore and bravado." Instead, he saw tough guys modeling the outcome of possible scenarios, based on past behavior, with the aid of computers. It was "a boring exercise in mathematics and probability," he recalled.
That got Sylvester thinking about transaction-cost analysis. Reflecting on the SEALs’ approach to uncertainty, he thought: "I don’t want to rely on just myself. When I go up against another execution, how can I ensure my probability of success?" TCA, in Sylvester’s view, is a necessary tool, but hardly compelling. He wanted to make it more rewarding for traders.
Sylvester has been a trader for more than 15 years, most recently at Babson Capital Management, in Boston, where he worked from 2001 until 2008. Since early last year, he has run the 10-person trading desk at JPMorgan Asset Management, which executes orders for 28 managers with a range of investment styles. The firm manages more than $1 trillion in total assets.
After mulling over how to improve trading outcomes, Sylvester hit upon the notion that traders should focus less exclusively on how the particular stock they’re given is trading and more on who they’re trading for. In other words, traders need to trade the portfolio manager, not the market.
So what this means is that traders can model the behavior of stocks for a period of time before and after a portfolio manager adds or subtracts them from a portfolio, and can use that past aggregate data to figure out how to trade stocks better for that manager. The premise is that portfolio managers typically have consistent reasons for buying or selling a stock, and that those reasons persist over time and over a large number of investment decisions. Importantly, the immediate post-trade stock behavior associated with those decisions can also reliably be projected into the future.
"The PMs have a stable investment process," Sylvester said. "We’re the ones who are more mysterious. We add chaos into the investment process by not intersecting properly with the PM’s investment decision." JPMorgan Asset Management worked with Abel/Noser Corp., its longtime provider of transaction-cost analysis data, to come up with a way to implement trading strategies that build on a manager’s history. Abel/Noser, which provides TCA to about 500 buyside customers globally, has worked with the bank’s asset management division since 2000.
Painting a Picture
With input from JPMorgan, Abel/Noser built visualization tools that cue the trader about "who the manager is" in terms of his or her investment decisions. The manager’s decisions are represented by various shapes, based on the average pattern of a stock’s price before and after the trading decision. JPMorgan usually looks at a stock from two days before the investment decision to two days after, but can change that period based on the stock’s characteristics.
"This paints a picture," said Peter Weiler, executive vice president of global sales at Abel/Noser. "In the past, we looked at the shape of a single stock, how traders executed it, what the volume was and what was happening in that name. Now we can do it by portfolio manager, on an aggregate basis, based on their past investment decisions."
In his view, this ties execution decisions more closely to the portfolio manager’s process when buying or selling the particular name. "It’s really creating a pre-trade estimate based on post-trade analysis of a PM’s trading decisions," Weiler said. "Traders can fine-tune their tactics to try to create a strategy consistent with that shape, which itself is remarkably consistent over time."
Andy Schwartz, a senior vice president at Abel/Noser, noted that clients frequently ask whether they should use a volume-weighted average price, strike price, percentage-of-trading test or some other metric to assess their trading costs. "The beauty and simplicity of these charts is that they cut through reams of numbers and answer that fundamental question very clearly," he said. "The minute a trader looks at a portfolio manager’s chart, he knows whether to be more aggressive or passive, and which strategy is likelier to produce better trading results."
Manager Algorithms
JPMorgan instituted this approach last summer for a handful of its portfolio managers. For those PMs, the desk created monthly "playbooks" to give traders a better sense of how each PM’s names usually fare after the investment decision is made. The desk is also developing manager-specific algorithms for traders to use. The goal is to automate some of the execution process that’s most suitable to a particular manager’s strategy under normal circumstances.
The JPMorgan portfolio managers that are part of this process are onboard with what the trading desk is trying to do. Sylvester noted that two portfolio managers who saw the trading patterns associated with their investment decisions immediately recognized them. The result has been a drop in trading costs.
Still, the portfolio-manager-based approach to trading is not an effective solution for all trades, Sylvester said. Small-cap stocks are less likely to follow similar patterns as consistently as large caps, he said. He added that "name-specific behavior" is hard to model. If the amount of anticipated alpha differs significantly for a portfolio manager on a name-by-name basis, then basing trading decisions on aggregate results is less likely to work.
However, JPMorgan’s efforts also offer the trading desk another benefit: Focusing on how a particular portfolio manager’s investment decisions should be traded can help the trading desk gauge the effectiveness of a manager’s instructions. For instance, a manager might tell a trader to get a trade on the tape urgently. But based on past experience, it might be better to execute the order more slowly to capture the mean reversion typically associated with that manager’s decisions.
While JPMorgan continues to expand its program, Abel/Noser is talking to other clients about the portfolio manager approach to execution and about creating "signature algorithms" based on portfolio managers, Weiler said. JPMorgan’s Sylvester said he’d be glad if other institutions look at this type of manager-focused trading analysis and incorporate it into their execution strategies. "If this helps, that’s great," he said. "It’s what we do with this information that’s proprietary."
(c) 2010 Traders Magazine and SourceMedia, Inc. All Rights Reserved.
http://www.tradersmagazine.com http://www.sourcemedia.com/
