Goldman Sachs introducing æalgorithm of algorithmsÆ
As more hedge funds and institutions use algorithmic trading strategies, choice is becoming as much a challenge as a boon.
Hedge funds and institutional investors in Asia are beginning to employ algorithms in their investment strategies, but the complexity can prove overwhelming to some. Two Goldman Sachs executives, Gene Reilly, managing director and head of pan-Asia equity trading and New York-based George Sofianos, vice president of equity execution strategies, explain how they are trying to bring simplicity to an increasingly complex set of trading methodologies.
How have fees for algorithms changed in the last year? Have they fallen?
Gene Reilly: IÆm not the first person to coin the phrase, but itÆs a ærace to zeroÆ for algorithms and direct market access (DMA). The real competition is the potential improvements in execution and the quality of the advice that a firm can provide. With unbundling, people may execute with fewer counterparties, but from a research perspective, speak to a wider number. This highlights why securities firms need to be able to offer leading algorithms and execution.
The explicit commission that clients pay is only a small part of the overall cost. A much larger percentage of the cost is the market impact of the missed opportunities, the speed and efficiency with which people are actually able to execute. So there is a huge focus on reducing those implied costs and the difference they can make to performance. The difference between high-quality execution and poor execution can cost funds returns of 200 basis points over the course of a year. ThatÆs material and Asian funds realize that strong execution is of paramount importance.
George Sofianos: We define execution into two areas: æhigh touchÆ and ælow touchÆ. By that we differentiate by commissions paid for the execution, whether say by direct market access or algorithm through to the full service broker/dealer suite of services. Increasingly there is a shift to the low touch.
What has been the evolution of algorithms?
Sofianos: Historically, there was a lot of experimentation in algorithms in the United States in the initial stages of introduction. In Asia you are now getting the benefit of the second and third generation of algorithms that have already been tested in the US.
Major broker/dealers offer a suite of around 12 algorithms. We offer a package of nine. Sometimes, being confronted with too many algorithms causes confusion with clients, hence the need for a consulting service.
The Volume Weighted Average Price is the granddaddy of algorithms. ItÆs straightforward and helps you with the volume profile of the trading day and slicing up an order accordingly, executing the order to mimic that profile.
The next evolution was algorithms designed to capture the spread for small orders. ItÆs called the small order spread capture algorithms. If you are patient and willing to wait a few minutes there is no reason for you to pay the spread, so letÆs have a strategy to capture that spread for the client. We call this the piccolo algorithm.
Some clients find that if they buy when the market is rising, the VWAP algorithm may not be a good execution strategy because it is way too passive.
So this led to the development of a second generation of algorithms, the so-called æshortfallÆ algorithms. In our case, our single-stock shortfall algorithm is called Forecast. This allows the user to calibrate the execution depending on how aggressive they want to be. The client chooses from 10 levels of aggressiveness and given that level then the algorithm optimizes the execution steps. 10: most aggressive. One: most passive. If you choose 10, you are going to frontload your execution and trade earlier.
Developments from this are ædynamic scalingÆ and ærelative scalingÆ. These are two more algorithms that evolved from the shortfall algorithm.
What is the main focus now?
Sofianos: Ours is from the single-stock trading to the portfolio algorithms. This is the third generation. So now we have a portfolio version of a shortfall algorithm. We call this the Portex algorithm. Once again a client may choose between æpassiveÆ, æmediumÆ or æaggressiveÆ and then the algorithm optimizes at a portfolio level. It is more complex now because the client may want to impose on the algorithm a set of restrictions on how to execute a basket of stocks. For example, if you want to be market neutral, you donÆt want your buys to get out of synch with your sells. This increases the complexity of the algorithm, so itÆs really the third generation.
DoesnÆt this terribly confuse some clients?
Sofianos: Responding to concerns from clients about the number of algorithms and the confusion this creates, we are now experimenting with an æalgorithm of algorithmsÆ. We call this the Navigator. Having taken the information from the user, it will decide itself which algorithm to use. It is a wrapper around the other algorithms, and the client just interfaces with Navigator with everything else done behind the scenes.
Are Asian funds using the same generation of products as are currently in use in the US?
Reilly: The tools available in Asia are just as sophisticated as those available to US clients. There is algorithmic access to the Japanese, Singapore, Korean, Hong Kong and Australian markets. In the near future there will be algorithmic access to ChinaÆs A-share market. In Asia we are providing exactly the same tools, logic and intellectual horsepower as we do in the States.
Do Asian funds trade in sufficient volume to make using algorithms worthwhile?
Reilly: Sure, absolutely they do. Regional market volumes are now much more significant and require that people use sophisticated tools to get the best execution.
What if you have fat fingers and press the wrong buttons when trading via an algorithm?
Reilly: WeÆve built in sophisticated screening into all of our algorithms, we work out limits, so that if the client enters something that seems too large, our monitoring team will hold on to the order, contact the client and say, ôThis is quite a large order, we just want to confirm it with you.ö If a trade is within the pre-determined boundaries, then it goes through almost instantly. We donÆt yet have a crossing-engine in Asia but are planning one for the near future û starting in Japan and afterwards in the rest of Asia.
Sofianos: YouÆre forced to think about order size, stock liquidity, whether small cap or large cap and degree of urgency. Depending on where you place your order in this matrix there is a right algorithm: a spread-capture algorithm for a low-urgency or small order in a liquid stock, or perhaps a VWAMP algorithm for a more urgent trade. If itÆs really urgent IÆd recommend a shortfall forecast algorithm.
We are working to develop a logic so that with the minimum specific indications of the clientÆs intentions, the Navigator will decide what the appropriate algorithm is going to be.
Where is the innovation?
Reilly: Our big clients do portfolio trades, portfolio risk trades, agency trades, single-share trades. They can request risk bids or offer for their single share trades. They can execute their order by direct execution, or they can use any of the suite of 10 algorithms. So they want to know how to think about all this and put in place a decision-making framework around choosing which of these execution channels to use.
What is new about our work is that weÆre developing a theoretical framework to enable clients and help them to make the right decision. Sometimes the instructions on the side of the tin arenÆt available from a bigger picture perspective.
Sofianos: WeÆre providing this as a consulting service. The client wants to be able to quantify the value-added by using these products compared to any trading costs and fees that he pays. Our research gives us insight into high-touch and low-touch executions and that research suggests that as the order of difficulty in an order increases, the high touch provides a better service and lower implicit costs.
Some clients must wonder if they could do a good or better job than the algorithm.
Reilly: In cases where the algorithm can do a better job, thatÆs when orders ought to be sent to the algorithm. Clients can then focus on the special-situations where they need to be focused by taking those other trades æoff their padÆ, so to speak. An algorithm is a productivity tool; it helps a buy-side desk trade efficiently, but donÆt make you a better trader.
How have fees for algorithms changed in the last year? Have they fallen?
Gene Reilly: IÆm not the first person to coin the phrase, but itÆs a ærace to zeroÆ for algorithms and direct market access (DMA). The real competition is the potential improvements in execution and the quality of the advice that a firm can provide. With unbundling, people may execute with fewer counterparties, but from a research perspective, speak to a wider number. This highlights why securities firms need to be able to offer leading algorithms and execution.
The explicit commission that clients pay is only a small part of the overall cost. A much larger percentage of the cost is the market impact of the missed opportunities, the speed and efficiency with which people are actually able to execute. So there is a huge focus on reducing those implied costs and the difference they can make to performance. The difference between high-quality execution and poor execution can cost funds returns of 200 basis points over the course of a year. ThatÆs material and Asian funds realize that strong execution is of paramount importance.
George Sofianos: We define execution into two areas: æhigh touchÆ and ælow touchÆ. By that we differentiate by commissions paid for the execution, whether say by direct market access or algorithm through to the full service broker/dealer suite of services. Increasingly there is a shift to the low touch.
What has been the evolution of algorithms?
Sofianos: Historically, there was a lot of experimentation in algorithms in the United States in the initial stages of introduction. In Asia you are now getting the benefit of the second and third generation of algorithms that have already been tested in the US.
Major broker/dealers offer a suite of around 12 algorithms. We offer a package of nine. Sometimes, being confronted with too many algorithms causes confusion with clients, hence the need for a consulting service.
The Volume Weighted Average Price is the granddaddy of algorithms. ItÆs straightforward and helps you with the volume profile of the trading day and slicing up an order accordingly, executing the order to mimic that profile.
The next evolution was algorithms designed to capture the spread for small orders. ItÆs called the small order spread capture algorithms. If you are patient and willing to wait a few minutes there is no reason for you to pay the spread, so letÆs have a strategy to capture that spread for the client. We call this the piccolo algorithm.
Some clients find that if they buy when the market is rising, the VWAP algorithm may not be a good execution strategy because it is way too passive.
So this led to the development of a second generation of algorithms, the so-called æshortfallÆ algorithms. In our case, our single-stock shortfall algorithm is called Forecast. This allows the user to calibrate the execution depending on how aggressive they want to be. The client chooses from 10 levels of aggressiveness and given that level then the algorithm optimizes the execution steps. 10: most aggressive. One: most passive. If you choose 10, you are going to frontload your execution and trade earlier.
Developments from this are ædynamic scalingÆ and ærelative scalingÆ. These are two more algorithms that evolved from the shortfall algorithm.
What is the main focus now?
Sofianos: Ours is from the single-stock trading to the portfolio algorithms. This is the third generation. So now we have a portfolio version of a shortfall algorithm. We call this the Portex algorithm. Once again a client may choose between æpassiveÆ, æmediumÆ or æaggressiveÆ and then the algorithm optimizes at a portfolio level. It is more complex now because the client may want to impose on the algorithm a set of restrictions on how to execute a basket of stocks. For example, if you want to be market neutral, you donÆt want your buys to get out of synch with your sells. This increases the complexity of the algorithm, so itÆs really the third generation.
DoesnÆt this terribly confuse some clients?
Sofianos: Responding to concerns from clients about the number of algorithms and the confusion this creates, we are now experimenting with an æalgorithm of algorithmsÆ. We call this the Navigator. Having taken the information from the user, it will decide itself which algorithm to use. It is a wrapper around the other algorithms, and the client just interfaces with Navigator with everything else done behind the scenes.
Are Asian funds using the same generation of products as are currently in use in the US?
Reilly: The tools available in Asia are just as sophisticated as those available to US clients. There is algorithmic access to the Japanese, Singapore, Korean, Hong Kong and Australian markets. In the near future there will be algorithmic access to ChinaÆs A-share market. In Asia we are providing exactly the same tools, logic and intellectual horsepower as we do in the States.
Do Asian funds trade in sufficient volume to make using algorithms worthwhile?
Reilly: Sure, absolutely they do. Regional market volumes are now much more significant and require that people use sophisticated tools to get the best execution.
What if you have fat fingers and press the wrong buttons when trading via an algorithm?
Reilly: WeÆve built in sophisticated screening into all of our algorithms, we work out limits, so that if the client enters something that seems too large, our monitoring team will hold on to the order, contact the client and say, ôThis is quite a large order, we just want to confirm it with you.ö If a trade is within the pre-determined boundaries, then it goes through almost instantly. We donÆt yet have a crossing-engine in Asia but are planning one for the near future û starting in Japan and afterwards in the rest of Asia.
Sofianos: YouÆre forced to think about order size, stock liquidity, whether small cap or large cap and degree of urgency. Depending on where you place your order in this matrix there is a right algorithm: a spread-capture algorithm for a low-urgency or small order in a liquid stock, or perhaps a VWAMP algorithm for a more urgent trade. If itÆs really urgent IÆd recommend a shortfall forecast algorithm.
We are working to develop a logic so that with the minimum specific indications of the clientÆs intentions, the Navigator will decide what the appropriate algorithm is going to be.
Where is the innovation?
Reilly: Our big clients do portfolio trades, portfolio risk trades, agency trades, single-share trades. They can request risk bids or offer for their single share trades. They can execute their order by direct execution, or they can use any of the suite of 10 algorithms. So they want to know how to think about all this and put in place a decision-making framework around choosing which of these execution channels to use.
What is new about our work is that weÆre developing a theoretical framework to enable clients and help them to make the right decision. Sometimes the instructions on the side of the tin arenÆt available from a bigger picture perspective.
Sofianos: WeÆre providing this as a consulting service. The client wants to be able to quantify the value-added by using these products compared to any trading costs and fees that he pays. Our research gives us insight into high-touch and low-touch executions and that research suggests that as the order of difficulty in an order increases, the high touch provides a better service and lower implicit costs.
Some clients must wonder if they could do a good or better job than the algorithm.
Reilly: In cases where the algorithm can do a better job, thatÆs when orders ought to be sent to the algorithm. Clients can then focus on the special-situations where they need to be focused by taking those other trades æoff their padÆ, so to speak. An algorithm is a productivity tool; it helps a buy-side desk trade efficiently, but donÆt make you a better trader.
¬ Haymarket Media Limited. All rights reserved.