BizAnalyst Network

Traders test the limits of technology in down markets

Traders discover the fallibility of technology
As computer-driven funds plummeted during August's market meltdown, algorithms are being scorned and analyzed. "Algorithms are extremely fallible; they are based on models of the universe but they cannot predict when the universe will stop working," said Michael Rosen of UNX. "The models failed because they could not conceive the perfect storm -- the model stopped conforming to the reality."

Traders test the limits of technology

full article

Computer-driven models are not the answer to the investment universe and everything
 
Algorithmic trading has been given a bad press recently, particularly after computer-driven funds lost the plot – or bought the farm – during last month’s market meltdown.

These quantitative funds – from previously gold-standard Renaissance Technologies to hedge funds within banks such as Goldman Sachs and Bear Stearns – reported losses as stock markets plunged.

It appeared everyone’s algorithms had the same idea and they all traded the same way, creating more volatility and liquidity hiccups.

So was last month the death knell of algorithmic trading? Could the love affair with black boxes be ending?

Tony Huck, a managing director at Investment Technologies Group, which launched a new type of algorithm to help manage portfolio risk, said: “It was the investment style that was to blame, not static algorithms. The traders have a lot of control – they can turn them off or on.”

The trick to making money with a computer-driven equity trading model is to ensure every potential market-related contingency is included in the algorithm before it trades. Unfortunately, that is not always possible.

Michael Rosen, vice-president for product development at agency brokerage UNX, said: “Algorithms are extremely fallible; they are based on models of the universe but they cannot predict when the universe will stop working. The models failed because they could not conceive the perfect storm – the model stopped conforming to the reality.”

If a fund uses a more portfolio-based algorithm, it could help it to better control what it does in an event-driven environment.

Sang Lee, founder and managing partner at consultancy Aite Group, said: “The first generation algorithms focused on single-stock trading and then moved on to facilitating portfolio-based trading with built-in parameters and flexibility, so client orders can respond to real-time changing market conditions.”

ITG’s latest algorithm, dynamic implementation shortfall, is a list-based algorithm, which means it encompasses a fund’s portfolio, not just a single stock. Others have similar algorithms – Goldman Sachs has one and Credit Suisse’s advanced execution services has one in its portfolio hedging device product line.

The benefit of using this type of strategy is the algorithm reacts in real time to benchmarks, spread levels, volatility and/or liquidity to execute the portfolio over single or multiple days. Huck said it is ideal for managing transitions over multiple days and can also be used on a daily basis for rebalancing, redemptions and new funds.

He said: “It is dynamic when looking at current market conditions such as volatility, volume or momentum. The user can change the trading parameters in real time if certain parameters are not being met.”

While newer types of algorithms, such ITG’s dynamic implementation shortfall, are helping traders to manage electronic trades and risk, there remain doubts about computer-driven funds.

Chris Martins, principal product marketing manager at algorithm trading platform vendor Progress Apama, said: “The perfect algorithm is never in play. The concerns raised with automated strategies and the markets they operate in was that when they go out of bounds of their expectations, they behave in problematic ways. That may be the danger in packaged algorithms.”

Because of the sub-prime induced bumps in the quantitative road, it may become less common for quant desks to allow an algorithmic strategy to run freely, without intervention. The ability to jump in and change – or stop – automated trading is critical.

Ary Khatchikian, chief technology officer of Portware, a supplier of automated portfolio trading software, said: “Some quants say they can handle it all electronically but you never know. That’s like saying there are no bugs in software.”

One benefit of a sophisticated algorithmic trading platform is the ability to back-test market data to see what went wrong and try different strategies that might work better in similar conditions.

Martins said: “Back testing is key. Funds can capture the volatile market data of early August and back test it in new strategies to see how they would react.”

But Rosen said predicting the future based on the past is not reliable. He said: “Algorithms are not prophets. This is not a critique of algorithms, it is a critique of the naivete of the market. They are ascribing powers of prophecy to algorithms, giving them an aura of power they were not built to have.”

Algorithmic trading is here to stay, despite the recent scares. The ability of vendors to tweak and improve their products should make quant traders feel better about using them.

Lee said: “I think something like ITG’s new algorithm will help firms react faster to changing market conditions but I don’t believe a full prevention is possible. Perhaps minimizing damage is more like it.”

Strategy that was blamed for escalating volatility spikes

Algorithmic trading

Algorithmic trading, also known as program or black-box trading, is a well-established technique on the buyside and accounts for about half of electronic trades in the mature stock markets.

It involves the routing of orders from fund managers’ trading blotter to specific computer-based algorithms that automatically manage the execution of these trades based on criteria such as timing, price or size of the order.

The first generation of trading algorithms, including volume-weighted average price and implementation shortfall algorithms, have been around for about three years but have come in for criticism this year for contributing to market volatility in periods of frenzied trading activity.

Algorithmic trading was blamed for escalating volatility spikes last month as the world’s leading equity markets reacted to sub-prime mortgage losses in the US.

Event-driven or smart algorithms

These are marketed as the next generation of smart algorithms, differing from the established tools in that they adapt in real time to market changes.

They emulate sellside traders by reacting immediately to the market and trading opportunistically to minimize execution costs and maximize returns, their vendors claim.

Customers can define the actions of the algorithms by setting parameters that optimize their performance while ensuring the underlying funds are not exposed to undue risk in times of extreme volatility.

full article