A while back I was asked the question “what do you see in the short-to-medium term so far as algos in Asia are concerned?” While Asia with its quirks and idiosyncrasies has meant that the standard set of algorithms (VWAP, TWAP, participation, plus an aggressive (ie will actively cross the spread) and a passive offering) have had to be all but re-coded for the various Asian markets, I felt that the average algorithm user (in many cases brokers trading on behalf of clients) would demand a rationalization of the algos available. At the time the company I was working for had just developed a fancy new algo and were having all the usual troubles that come with deploying it in Asia without any testing whatsoever. Other companies were promoting the fact that they offered over 150 different algorithms across several dozen brokers.
My prediction was based on the 80-20 rule (and let’s face it, the lion’s share of the volume from funds trading Asia via algorithms do so via VWAP), and has been somewhat correct, with one major broker that I know of decommissioning two algorithms, probably due to the cost-ineffectiveness of maintaining rarely-used algorithms in the region. However today I see on Asia Etrading that the march continues, with Instinet adding one more algorithm to their suite.
This news reminded me of some of the challenges of supporting algorithms in Asia which I have noticed over the last few years.
Local Development Team
For starters, you need to have a team locally who knows the intricacies of the various Asian markets, and is available for support during Asian hours. No longer is it acceptable to raise issues to the US support team, and get back to the client the next day. If you’re not prepared to commit to a local development team then you’re probably not going to be taken seriously by clients.
One unavoidable consequence of this is that upgrades branch entirely from their US/Europe upgrade path. And when you are already thin on resources in the area you can find you have an ever-increasing gap between client demands and the implemented reality. Especially for clients trading Asian markets from the US or Europe, all they want is for their algorithms to function the same, seamlessly across regions. If you’re involved in the support of these clients be prepared to explain to the folks in Sales and to the client directly. I recommend a proactive approach, but I understand that this isn’t always feasible.
A Few Good Men
Related to the above point, sourcing support staff can also be an issue. You can’t just fly someone over from the US, unfamiliar with local markets, in order to oversee a server-side upgrade of an algo. This also affects emergency replacement of staff when people leave – for other roles you can often easily just substitute someone in from another region temporarily to cover until a replacement is found (expect a post from me talking about this in more detail shortly).
Infrastructure in Every Region
As algorithms tend to be implemented as a separate service running on a server, it is possible to co-locate just the algorithm server in certain cases. Ironically the firms that can support this infrastructure best are large, established operators, whereas it is the smaller firms that could really benefit from being able to separate the infrastructure. For comparison, the “smaller” firm I worked at previously chose to have the majority of the infrastructure in Hong Kong. TSE-bound orders were sent out from Hong Kong “post-processing” to a broker in Japan. Needless to say, latency was a significant issue.
All-in-all the issues for supporting algorithms in Asia are similar to the issues for supporting other electronic trading support functions. However I find that the issues tend to be more pronounced, and the severity of mistakes can be much higher. I guess on the flip side, for those companies who are prepared to commit to providing and supporting algorithms properly, localizing and fine-tuning them as appropriate, it can definitely give the impression that the company has a strong commitment to the region.