Exchange Views: ASX - Ramping Up For Algo/DMA Trading
In common with other exchanges around the globe, the ASX has seen algorithmic and automated trading fundamentally change the nature of its order flows. AT talks to David Stocken, Senior Product Manager, Equity Products at the ASX about how the exchange is responding to this fluid environment.
How has algo/DMA trading grown as a percentage of the ASX's activity?
DMA/algo trading now accounts for approximately 25% of turnover
by value on the ASX. This is up from approximately 20% about
twelve months ago. Over the last eighteen months we have been
working on several initiatives to enhance DMA and algorithmic
- The introduction of anonymous trading in November 2005, which has helped institutions reduce their footprint and therefore also their market impact.
- The introduction of a FIX interface to our trading platform.
- Moving to a new trading platform (ITS), which combines lower latency and greater capacity.
- A new fee structure (based more on value rather than volume) for brokers. All of these initiatives have helped facilitate DMA/algo trading on the exchange. How high will these volumes go? I am sure we will continue to follow the international trend upwards. All end users that I speak to plan to increase their electronic trading activity in coming years.
In an interview in July 2005, former ASX CEO Tony D'Aloisio said that algorithms were being used on the exchange by the full range of institutional investors (pension funds, hedge funds and investment banks etc). Is activity pretty evenly balanced or does one group predominate?
I would agree that algorithms are being used by a full range of institutions. Some hedge funds may simply prefer to auto-execute and therefore prefer DMA. Other hedge funds may trade off technical signals and then prefer to use an algorithm to execute their stock. Several of the Australian fund managers I speak to are also using algorithms. So I would say that algorithmic usage is pretty evenly balanced across a wide range of users.
In the same interview he said that most ASX DMA business comes
from offshore. Is that still the case? Do you expect the balance
to change and domestic DMA activity to grow?
There is no doubt that a large proportion of our DMA activity comes from off-shore. Offshore funds have probably been exposed to algorithms for slightly longer than our domestic fund managers and therefore feel more comfortable with them. However I would have to give our domestic funds industry high marks for quickly learning about the benefits of algorithmic trading. In a short space of time our domestic funds have really increased their knowledge and usage of algorithms. But I would expect both offshore and onshore fund managers to increase their DMA activity in coming years.
ASX trade count and average trade size
Who do you see as the customer category with the greatest potential for algorithmic trading growth?
I would say Asian hedge fund sector would be a strong potential candidate for this title - especially since it is expected to show very substantial general growth over the next few years.
What effect have the ASX pricing changes had on capturing algo
The new pricing for brokers came into effect on July 1 2006. Basically we moved away from our old volume pricing structure towards a value pricing structure that gives the DMA/algo broker more pricing certainty.
The immediate effect was a dramatic drop in average trade size in July, down to about $25000. However since then the average trade size has levelled off but is still lower than under the old pricing system. This implies that more algorithmic activity has occurred since our pricing change. Our trade count has also risen aggressively. However it must be said that there are many other factors behind this, including the current strong bull equity market.
What is the current position with regard to the proposed changes
to ASX crossing rules?
The exchange is reviewing a number of its crossing rules in response to market feedback. All proposed changes need to be discussed with the regulator, ASIC (the Australian Securities and Investments Commission). At this stage, we're not in a position to outline what those changes might be.
What have been the main changes to the ASX trading platform(s) as
regards connectivity and throughput?
Recent improvements to connectivity and throughput have come through the implementation of the Integrated Trading System (ITS). All products that were previously traded through two trading systems (SEATS and CLICK) are now available on a single platform and therefore a single connection/ API. We are also now able to offer contingent trading across multiple product types.
What technology is the ITS infrastructure based upon?
The ITS system consists of a cluster of five OpenVMS systems spread across two locations. This is then connected via ASX's fully redundant private network to gateways (Windows servers) based at participant sites. For overseas participants and for participants who don't wish to house the gateways on their premises we also offer a hosting facility where the gateway is located on ASX premises and the participant connects to the gateway via a VPN connection.
ITS replaced the SEATS mainframe based system, which was some twenty years old. In common with other exchanges, we have seen message volumes increase substantially on the back of DMA and algorithmic trading.
The ASX then... and now
Do you provide a Colocation service so that customers can run their algorithms on servers based at the exchange?
At present we don't provide this particular service, although as I mentioned earlier we do offer locally hosted exchange gateways. However, we are aware that there is customer demand for Colocation and that other exchanges provide it, so we are currently reviewing this.
What sort of performance levels can ITS achieve?
ITS has been built to handle large volumes and to provide fast order transaction capabilities. We have successfully tested it to over 4,000 order book changes per second sustained, for peaks of over 8,500 order book changes per second, over 200 trades per second sustained, 500,000 trades per day and 60 million order book changes per day. Order latency has reduced by 3-4 times since the move from SEATS to ITS. Typically entering an order and receiving an acknowledgment takes around 30 milliseconds (including network latency) but can be as low as 15 milliseconds.
In addition to performance gains, ITS also introduced a new transaction type particularly suited to algorithmic and DMA trading. This transaction type allows the entry of multiple orders in a single transaction.
What is the exchange's strategy as regards FIX?
We have two FIX services, both of which are live. The first is the Cameron Universal Server. The adaptation required for this software to connect to ITS was introduced in partnership with Cameron Systems. The product supports FIX versions 4.0 - 4.2. We chose this product because of Cameron Systems' reputation for having the lowest latency FIX engine and we have not regretted that decision. We currently have a small number of participants who connect this way but this is already growing. The Cameron Universal Server is also available as a separate product from Cameron Systems and we have a few participants who have chosen to go this route, as Cameron offers a number of additional features to allow easy integration with a participant's internal systems.
The second FIX service is the FIX market data service also provided by Cameron Systems. This provides real-time market information in FIX format. We have a number of participants and data vendors who receive market data in this form. Certainly for international participants and data vendors it is easier and cheaper to maintain systems that talk FIX to multiple exchanges than to have to connect using each exchange's proprietary API.
Furthermore, the ASX is a member of the FIX Protocol Exchanges Working Group for Asia Pacific. We use our membership of this group to keep abreast of FIX developments.
What is the ASX view on the FIX FAST protocol? Will the exchange
be adopting it?
In terms of FAST, we are likely to upgrade to this for market data. The standard is still very new so we will wait for more up take before proceeding. FAST's reduction in bandwidth utilisation is certainly attractive, as it will reduce the cost of disseminating data whilst simultaneously increasing the speed.
Are you starting to see growth in algo trading of futures as well
Yes, algorithmic futures execution is definitely expanding. The recent merger of ASX and the Sydney Futures Exchange (SFE) is very interesting in the context of algorithmic execution expanding beyond equities to other asset classes, such as futures. Futures algos are more about opportunity discovery rather than the finessing of order execution often associated with equities. On the SFE algos are probably used more to identify trading opportunities based on complex combinations of circumstances, eg, spread trading opportunities between futures/futures, futures/cash, currency movements and all other variations of the same theme, or alternatively high frequency trades based on changes in technical analysis indicators or cutting edge quant theory. It is also worth pointing out that the SFE's trading platform SYCOM is FIX based. As a newly combined entity, the ASX will look to bring SYCOM into line with current FIX standards thereby ensuring a consistency between ITS and SYCOM.
Is high frequency automated trading of futures already
commonplace on ASX/SFE?
Absolutely; and furthermore we are increasingly seeing more and more end customers demanding direct access beyond that offered by the brokers based on latency, order management and algo requirements.
Have you seen much evidence of multi asset algorithmic trading as
There has been a lot of talk about this at conferences of late, but as yet we haven't had customers approaching us about it. However, it seems inevitable that it will eventually take off.
Do you see algorithmic trading of options as the next evolution
for the ASX/SFE?
This is a fascinating development in the algorithmic evolution. I was formerly an options market maker for many years and am heavily involved with the development of our Exchange Traded Options market at ASX. I have been reading with great interest about the progress that some brokers are making developing algorithms for ETO execution. While I understand that these algorithms are more complex to develop than regular equity algorithms I think their development is very important.
Despite having a healthy ETO market with good liquidity in our top tier of stocks we still have a lack of "patience" when it comes to dealing ETOs in size. For example, an institutional dealer will typically prefer a broker to execute their line of, say, 1000 options in one print so that the institutional dealer can move onto his next item of business. Unfortunately there is market impact when dealing 1000 options in one line. Naturally the market maker wants to factor in some comfort factor for dealing size.
However, the institutional dealer should have a lower market impact if he deals his 1000 options in, say, 100 blocks of 10 steadily over the day. ASX has a very sophisticated ETO market. We were one of the first markets to introduce tailor-made combinations and have a high usage of derived bait orders. I think institutional dealers would be pleasantly surprised about their ability to successfully deal ETOs via an algorithm with lower market impact, particularly on the liquid ETO stocks at the ASX. Just as the buyside agrees that market impact on equities can be reduced via the use of an algorithm, the same should hold true for ETOs. To that end we would welcome brokers looking to introduce algorithmic ETO execution on ASX.