The Gateway to Algorithmic and Automated Trading

A basket of abstractions

Published in Automated Trader Magazine Issue 10 Q3 2008

First person in the Q3 issue is South Africa-based subscriber Schalk Myburgh, Business Analyst, Nedbank Equity Capital Markets, who shares his experience of automated and algorithmic trading on the Johannesburg Stock Exchange with editor William Essex.

William Essex talks to Schalk Myburgh

William Essex: Please discuss your firm's take-up of algorithmic and automated trading functionality.

Schalk Myburgh: We recently acquired a Complex Event Processor (CEP) that we integrated into the Johannesburg Stock Exchange (JSE) enabling us to analyse thousands of messages per second. The CEP also has a direct connection into the JSE resulting in quick reaction to changes in the market. We have used the CEP to implement an automated market maker that calculates about 200,000 prices and sends about 50,000 orders to the market per day. Currently we are developing a basket trading infrastructure to enable our current proprietary trading activities to benefit from the increased speed and functionality of the CEP. For the future we will be interacting with the market more electronically including statistical arbitrage and algorithmic trading.

William Essex: Will this continue, and perhaps accelerate? How will it, or has it, determined, or indeed altered, the configuration of your trading function?

Schalk Myburgh: Auto/Algo trading will definitely grow and become more important in the future. We see it as an abstraction of trading where a person interacts with the market at one step removed.
So instead of sending manual orders to the market all day long, the trader must now be able to guide the algorithm in the direction it needs to go. This abstraction can be applied to any type of electronic trading, whether it is an algorithm trying to achieve best execution or if it is doing statistical arbitrage to generate alpha. This abstraction will become more important in the future as it will be harder to be competitive if you are still doing everything hands on.

William Essex: To what extent, if at all, has electronic trading determined or altered your company's overall trading strategy?

Schalk Myburgh: Due to the high volume of orders, we would not have been able to compete in the market-making business without a fast, agile infrastructure to interact in the market. We wish to become more active in the algorithmic trading arena as a result of this.

William Essex: To what extent is the take-up of such functionality changing the trading environment in South Africa?

Schalk Myburgh: We have seen an increase in the number of orders routed to the exchange and are seeing more automated trading activity on the order books. One major factor to consider while implementing algorithms in South Africa is that our orders get routed to London where the central order book resides. This means that each order must be routed to London and acknowledgement of the order must be routed back to South Africa. This imposes a round trip latency of about 150 milliseconds. This latency cannot be reduced as it is limited by the speed of light and the distance to be travelled. This means that if you have sent an order to the market you need to wait at least 150 milliseconds before you can modify the order. In the algorithmic trading game this is a long time to wait if the market has changed in or out of your favour.

William Essex: What are the asset classes for which you use algorithmic and automated trading? And for what purposes?

Schalk Myburgh: We are market makers on equity warrants and ETFs and use automated trading on equities to generate alpha.

William Essex: What lessons, if any, do you draw from the US and European electronic trading experience?

Schalk Myburgh: We have closely monitored the development of technology in Europe and we have learned the following: first, algorithmic trading is a race to be first in line when an opportunity arises, in terms of design, implementation and latency; secondly, that it is important to be able to integrate algorithms quickly and generically onto clients' desks; and thirdly, that regulation will become stricter in terms of executing for the buyside. That said, we only have one execution venue, so for us to prove we tried to achieve best execution is a lot easier than it is in Europe under MiFID. Also, we have learned that the buyside is becoming more active in measuring the performance of their brokers using post trade analysis.

William Essex: Does electronic trading simplify, complicate or otherwise alter the oversight process and management of risk?

Schalk Myburgh: Algorithmic trading adds operational risk to a trading strategy. As the algorithm has access to the market, while it is operational it must behave flawlessly otherwise it can result in large market impact or unwanted position build-up if not properly designed. Algorithms must be monitored intraday and metrics must be added to monitor if the algorithm is behaving as designed. You do not want to use the wrong price if you are executing 100 client orders on your order book. The only other change we see is that the volume of trades goes up as you execute algorithmically, so if your risk system can handle the volume there should be no difference in measuring market risk.