Wallstyouth
Vice President
- Joined
- 5/25/07
- Messages
- 116
- Points
- 28
Algorithmic trading gets smarter after quant upset
Sarfraz Thind 04 Oct 2007
Extreme events this summer exposed weaknesses in computer strategies
The knives are out in the ultra-secretive world of quant-based trading. The past two months have not been kind to computer-driven funds, including those run by Renaissance Technologies, DE Shaw, Goldman Sachs, JP Morgan's Highbridge Capital Management, US hedge fund Tykhe Capital, State Street Global Advisors and AQR Capital, which are among the many to have reportedly suffered billion-dollar losses.
Indeed, the algorithms that underpin the strategies of these funds - deleveraging their positions in the tightest of market conditions - have been blamed for contributing to the increased equity market volatility.
Jim Simons, founder of Renaissance Technologies, one of the most successful quant-driven funds in the world, in a letter to the firm's investors, said: "While we believe we have an excellent set of predictive signals, some of these are undoubtedly shared by a number of long/short hedge funds.
For one reason or another many of these funds have not been doing well, and certain factors have caused them to liquidate positions. In addition to poor performance, these factors may include losses in credit securities, excessive risk, margin calls and others."
David Viniar, chief financial officer of Goldman Sachs - whose flagship quantitative multi-strategy Global Alpha hedge fund is down 32.9% this year to mid-September, and in August plunged 22.5% - termed August's market moves as a 25-standard deviation event, something that would normally only occur once every 100,000 years. In September, the fund was valued at $6bn (€4.3bn), down from $10bn last year.
While managers have posited various explanations, few have blamed their own models for the losses. The growing popularity of quant strategies is partly to blame, since they do not necessarily require huge resources, just technical know-how and a computer.
This inevitably means funds are sharing similar positions and strategies. The nature of the high-frequency trades, which rely on speed to arbitrage minuscule mispricings, coupled with overcrowding in the market, means many funds are trying to escape the same trades at the same moment.
The difficulty is factoring in liquidity squeezes. It is not as easy as it looks, according to John Edge, European head of electronic client solutions at JP Morgan in London. He said: "One of the key things in this episode was the challenge of extreme events, which suddenly compromise previously performing models.
"Repeatedly adjusting a model and altering previously well-thought-of positions in such a short time frame, especially in leveraged positions, is a difficult and sometimes costly exercise."
Indeed many of the firms suffering in the recent crisis will point to the successes that their models have delivered since the beginning of the year as a reason not to over-react. Reassurances that opportunities exist despite the turbulence appear to be satisfying investors. Such is the confidence of investors that many have ploughed additional investment into the funds in recent weeks.
Goldman Sachs' $3bn cash injection for its Equity Opportunities fund, included $1bn of external investor money, while SAC Capital raised $1bn in August and AQR Capital Management secured $1bn last month.
But some tweaks to the underlying algorithms will be going on to ensure they are more sensitive to market patterns, according to John Bates, founder and vice-president of Apama Products at Progress Software. Smarter algorithms are being developed.
Bates said: "You need to have an algorithm that can deal with any circumstance, including extreme events. Many of the institutions haven't been prepared for the extreme events seen in recent months."
Bates expects quant trading models to increasingly incorporate risk rules alongside trading rules. This would allow, for example, an algorithm to automatically trade on a change in a particular risk measure, such as value at risk, as well as trading on arbitrage opportunities, said Bates.
He added: "We are seeing algorithmic trading spread across asset classes - from equities to futures and options and foreign exchange - as people look at cross-asset trading to hedge a position. We are also seeing real-time compliance being incorporated into the underlying model."
Further innovations will allow real-time news to be incorporated into the underlying algorithm, enabling a trading model to instantaneously react to developments in the market.
Bates said at the moment this was done by only a handful of the most sophisticated hedge funds. He pointed to a move by Apama to incorporate Dow Jones' news feed into its trading platform as a signal that such services might soon be embraced by the wider statistical arbitrage community.
Ary Khatchikian, president and chief technology officer at Portware, an algorithmic trade management system provider, said: "The same products and algorithms that are being used by the largest funds are also available to the smallest fund. It is a case of the survival of the fittest in this world. The ones that can find opportunities quicker than others will have the advantage."
Brokers have also been looking to develop algorithms that underpin order execution - a crucial element of any fund manager's strategy. Edge said: "If you can get a trade done without moving the price, that's the nirvana outcome which traders aspire to and which brokers are trying to provide the best tools to facilitate."
JP Morgan has implemented enhancements to its trade execution algorithms, which are designed to pick up on subtle patterns in the order book and adjust the algorithm's execution strategy in an instant.
He said: "There is a host of signals on the order book. If we build an algorithm to interpret these signals on an intra-trade basis then it gives the trader using these algorithms an advantage. We use the term harvesting liquidity for algorithms that are able to draw out and manage liquidity in this way."
US agency broker ITG has launched a new breed of smart trading algorithms, Dynamic Implementation Shortfall, which react to market events and trade accordingly.
It trades opportunistically, enabling fund managers and hedge funds to optimise profitability and lower trading costs, according to the broker.
Customers choose from a list of performance objectives and the system executes the strategy, generating live trades as markets move. The broker said Dynamic differed from implementation shortfall algorithms, which are static and follow a predetermined trading schedule.
The UK's Financial Services Authority is the first regulator to monitor market abuses using an algorithm. In June it entered into a partnership with Detica, a technology consulting firm, using Progress Apama's event-processing platform. The technology will enable the regulator to track possible cases of insider trading by analyzing unusual market trading patterns.
As dependence on algorithmic trading increases, strategies will need to grow up.
Although quant funds were not the only ones to take losses in this summer's crisis, they have learnt their lessons the hard way. Investors will be hoping that new breeds of algorithms will not repeat mistakes of the past.
Sarfraz Thind 04 Oct 2007
Extreme events this summer exposed weaknesses in computer strategies
The knives are out in the ultra-secretive world of quant-based trading. The past two months have not been kind to computer-driven funds, including those run by Renaissance Technologies, DE Shaw, Goldman Sachs, JP Morgan's Highbridge Capital Management, US hedge fund Tykhe Capital, State Street Global Advisors and AQR Capital, which are among the many to have reportedly suffered billion-dollar losses.
Indeed, the algorithms that underpin the strategies of these funds - deleveraging their positions in the tightest of market conditions - have been blamed for contributing to the increased equity market volatility.
Jim Simons, founder of Renaissance Technologies, one of the most successful quant-driven funds in the world, in a letter to the firm's investors, said: "While we believe we have an excellent set of predictive signals, some of these are undoubtedly shared by a number of long/short hedge funds.
For one reason or another many of these funds have not been doing well, and certain factors have caused them to liquidate positions. In addition to poor performance, these factors may include losses in credit securities, excessive risk, margin calls and others."
David Viniar, chief financial officer of Goldman Sachs - whose flagship quantitative multi-strategy Global Alpha hedge fund is down 32.9% this year to mid-September, and in August plunged 22.5% - termed August's market moves as a 25-standard deviation event, something that would normally only occur once every 100,000 years. In September, the fund was valued at $6bn (€4.3bn), down from $10bn last year.
While managers have posited various explanations, few have blamed their own models for the losses. The growing popularity of quant strategies is partly to blame, since they do not necessarily require huge resources, just technical know-how and a computer.
This inevitably means funds are sharing similar positions and strategies. The nature of the high-frequency trades, which rely on speed to arbitrage minuscule mispricings, coupled with overcrowding in the market, means many funds are trying to escape the same trades at the same moment.
The difficulty is factoring in liquidity squeezes. It is not as easy as it looks, according to John Edge, European head of electronic client solutions at JP Morgan in London. He said: "One of the key things in this episode was the challenge of extreme events, which suddenly compromise previously performing models.
"Repeatedly adjusting a model and altering previously well-thought-of positions in such a short time frame, especially in leveraged positions, is a difficult and sometimes costly exercise."
Indeed many of the firms suffering in the recent crisis will point to the successes that their models have delivered since the beginning of the year as a reason not to over-react. Reassurances that opportunities exist despite the turbulence appear to be satisfying investors. Such is the confidence of investors that many have ploughed additional investment into the funds in recent weeks.
Goldman Sachs' $3bn cash injection for its Equity Opportunities fund, included $1bn of external investor money, while SAC Capital raised $1bn in August and AQR Capital Management secured $1bn last month.
But some tweaks to the underlying algorithms will be going on to ensure they are more sensitive to market patterns, according to John Bates, founder and vice-president of Apama Products at Progress Software. Smarter algorithms are being developed.
Bates said: "You need to have an algorithm that can deal with any circumstance, including extreme events. Many of the institutions haven't been prepared for the extreme events seen in recent months."
Bates expects quant trading models to increasingly incorporate risk rules alongside trading rules. This would allow, for example, an algorithm to automatically trade on a change in a particular risk measure, such as value at risk, as well as trading on arbitrage opportunities, said Bates.
He added: "We are seeing algorithmic trading spread across asset classes - from equities to futures and options and foreign exchange - as people look at cross-asset trading to hedge a position. We are also seeing real-time compliance being incorporated into the underlying model."
Further innovations will allow real-time news to be incorporated into the underlying algorithm, enabling a trading model to instantaneously react to developments in the market.
Bates said at the moment this was done by only a handful of the most sophisticated hedge funds. He pointed to a move by Apama to incorporate Dow Jones' news feed into its trading platform as a signal that such services might soon be embraced by the wider statistical arbitrage community.
Ary Khatchikian, president and chief technology officer at Portware, an algorithmic trade management system provider, said: "The same products and algorithms that are being used by the largest funds are also available to the smallest fund. It is a case of the survival of the fittest in this world. The ones that can find opportunities quicker than others will have the advantage."
Brokers have also been looking to develop algorithms that underpin order execution - a crucial element of any fund manager's strategy. Edge said: "If you can get a trade done without moving the price, that's the nirvana outcome which traders aspire to and which brokers are trying to provide the best tools to facilitate."
JP Morgan has implemented enhancements to its trade execution algorithms, which are designed to pick up on subtle patterns in the order book and adjust the algorithm's execution strategy in an instant.
He said: "There is a host of signals on the order book. If we build an algorithm to interpret these signals on an intra-trade basis then it gives the trader using these algorithms an advantage. We use the term harvesting liquidity for algorithms that are able to draw out and manage liquidity in this way."
US agency broker ITG has launched a new breed of smart trading algorithms, Dynamic Implementation Shortfall, which react to market events and trade accordingly.
It trades opportunistically, enabling fund managers and hedge funds to optimise profitability and lower trading costs, according to the broker.
Customers choose from a list of performance objectives and the system executes the strategy, generating live trades as markets move. The broker said Dynamic differed from implementation shortfall algorithms, which are static and follow a predetermined trading schedule.
The UK's Financial Services Authority is the first regulator to monitor market abuses using an algorithm. In June it entered into a partnership with Detica, a technology consulting firm, using Progress Apama's event-processing platform. The technology will enable the regulator to track possible cases of insider trading by analyzing unusual market trading patterns.
As dependence on algorithmic trading increases, strategies will need to grow up.
Although quant funds were not the only ones to take losses in this summer's crisis, they have learnt their lessons the hard way. Investors will be hoping that new breeds of algorithms will not repeat mistakes of the past.
