Best of Breed in Global Algorithmic Trading by Gary Ardell & William Capuzzi Published by THE TRADE Algorithmic trading & smart order routing - 3rd edition
Best of breed in global algorithmic trading William Capuzzi and Gary Ardell* Algorithmic trading accounts for roughly 50% of buy-side order volume around the world. At least 60 brokers have some sort of algo offering. Between all these brokers, there are hundreds of algos. How do you choose the best algo brokers for you? Is there such a thing as best of breed? This chapter attempts to set out the factors by which any algorithm provider can be judged best of breed (or not). These factors, when combined, allow you to narrow your choices down to a manageable number of racks with which to work. We find six criteria that constitute a practical basis for judging best of breed in algorithmic racks: workflow, technology, service, differentiation, anonymity, and, of course, performance. Let s look at each one in turn. Workflow Individual algorithms from a particular provider may do a superb job in the right circumstances. However, it is important to look at the rack as a whole. It should be convenient and Performance Anonymity Differentiation Service Technology Workflow intuitive for your traders to use. It needs to provide the right tools and offer access to the right markets. Be sure the rack is customisable to your expectations. Start with the basics. How clean and intuitive is the broker s rack display? There are about 50 non-proprietary front ends used globally. Each one displays the algorithms differently. Your broker should ensure that their display works intuitively for your traders across all front ends. Your traders shouldn t need a worksheet to figure out what an algo does or what its tags mean: if they do, workflow is compromised. Your brokers should try to * William Capuzzi, managing director, global electronic trading Gary Ardell, managing director, financial engineering & advanced trading solutions ConvergEx 31
32 make their features intuitive and consistent for traders regardless of the front end without sacrificing the bells and whistles. Good workflow demands consistency. How, for example, does your broker s VWAP algorithm work in different regions? For a tool to be global, it has to function consistently in different markets. Of course, not all algorithms are designed for general use, but if they are, they should be engineered to reach the same ends across different venues. It is frustrating for a buyside trader using multiple providers to keep track of not only each provider s offering in aggregate, but also what works in which market within each rack. There are racks in which similar algorithms have different tags for different regions, and there are cases where certain tags do not work in some countries. Presentation is important. If you walk into a pharmacy to find the shelves dusty and disorganised, you might want to look elsewhere to fill your prescription, even if the medications they stock are perfectly fine. Many brokers end up with presentation inconsistencies because of the way they have grown. These brokers may have grown by acquisition and find that they now need to amalgamate two or three different algorithmic racks. The onus remains on the provider to overcome these challenges, not on your traders. Best of breed algorithmic providers roll out new algorithms or feature sets on a regular basis. Given the diversity of front ends, ergonomic design must be a priority. Best of breed brokers make sure their rack stays fresh without sacrificing ease of use and quality of workflow. Technology Three distinct technological areas define best of breed brokers: geographical optimisation; liquidity access, and the quantitative underpinnings of the algorithms themselves. Geographical optimisation requires a detailed knowledge of individual markets specificities. For example, the market open in the US is very different from the open in Australia, which is executed in alphabetical order. In another example, in Turkey, you cannot cancel a limit order, and in South Africa, you have unscheduled auctions that are event-driven. Unless all those details are accounted for, the algorithm won t work properly. Liquidity access isn t only increasingly important in smaller markets. Major markets have witnessed significant fragmentation in recent years. For example, in the US, it is now impossible to provide
a comprehensive electronic trading offering without efficient access to various dark pools. Best of breed brokers spend a lot of money to build a robust infrastructure capable of such access. In addition, algorithmic design needs to be consistent with trades conducted away from the primary market. In Europe, for example, the only time a ConvergEx algorithm will trade on an MTF is if the price is equal to or better than on the primary exchange. A VWAP in this case will always be better than one trading only against the primary exchange. A broker s quantitative infrastructure is a good indicator of the resources available not only to initiating, but also maintaining the strength of their algorithmic offering. Building a simple, effective VWAP engine is well within the scope of many firms, but as the range of strategies becomes broader, the more dedicated resources need to be. Service Nearly always, if you put the algorithm into the system and leave it to do its job, it will perform as you think it should. However, there is a level of comfort in knowing that on the provider side, an extra pair of eyes, what we call the electronic sales trader, is available and behind that is a human brain that understands the algorithm and how it is supposed to work. Maybe not every buy-side firm wants a second set of eyes, but the ability to provide the service if wanted puts a broker into the best of breed category. Service can be difficult to measure, but reputation tends to circulate by word of mouth, particularly in a period where economic turmoil has resulted in the downsizing of many sell-side trading desks. Differentiation It may seem odd to highlight the importance of differentiation after discussing the importance of consistency, but here we refer to the range of algorithms in a rack beyond the instantly recognisable ones. Even algorithms everyone believes they are familiar with, such as VWAP, Initiation Price, and POV, must be customisable to work in line with what each trader is trying to do. There is no one size fits all algorithm, even for the so-called mainstream strategies. Take VWAP perhaps everybody s first algorithm for example. Traders have differing views about how or whether the open and close ought to be included, as some markets are very efficient at the open and the close and others are random. The algorithm needs to be adaptable to take the trader s opinion into account. Suppose that a trader enters a limit price into a 33
34 VWAP algorithm and the market trades up to it. What happens when the market comes back? How should the algorithm treat that period when it was out and how should the volume be rescheduled for the day? Initiation price is another example. One trader might say, If the market starts to move away from me, become more aggressive; but if it starts to move towards me, become less aggressive. Others might want to be aggressive in both directions. Many highly effective traders have strongly differing opinions on these matters. Your broker s algorithms need to allow for your traders different approaches. A best of breed provider should also include unique and compelling offerings in their rack that are unavailable elsewhere, designed to fulfil your specific trading needs and address gaps in algorithmic trading as a whole. An example of an algorithm that fills such a gap is ConvergEx s ADR Plus SM. This algo allows clients to optimally source liquidity for their ADR trades in both overseas and US markets and to achieve better price execution without having to manage the ADR conversion process. It employs a user-defined, market-specific strategy with a dollar price limit and is available for both US exchange-listed ADRs and Pink Sheet traded ADRs. Even though part or all of the order will be traded overseas, the ADR Plus workflow permits clients to send US symbols from their OMS or EMS in the same way they would for a US traded order. Each algo slice will flow back in dollars to the client s OMS/EMS without manual intervention by the client. Anonymity Anonymity has produced a fair degree of scaremongering in the algorithmic trading world and is one of the major fault lines between full-service firms and agency providers. The latter imply that the former may leverage their prop desk off their algorithms, while the former go out of their way to demonstrate total segregation between prop and client flow. This is a subject that merits rational review. The client base of the major full service houses suggests that there is no shortage of buy-side firms that are fine with dealing with a prop shop. The old worries that an order phoned in to a sales desk might be overheard and picked up by a colleague are largely a thing of the past. However, even without looking into the orders, there are other information indicators (such as the trading direction of particular
sectors) that if used by the proprietary traders, may cause the value of the algorithm to degrade over time. Agency providers are free of this potential conflict. Performance At the end of the day, performance is what matters most. The challenge is how to assess it, especially for brokers you are just now thinking about bringing on. Looking at published reports clearly does not help. One broker is at the top of one list and at the bottom of another. The only performance that matters is the performance a broker delivers on your orders. There is only one sure-fire way to judge the comparative performance of different providers and that is through a bake-off. At ConvergEx, we typically ask new clients to give us a third of their business and then pit us against other viable best of breed candidates. After a couple of months, it should be clear who is best at what. Conclusion How many viable best of breed brokers are there? The answer depends on your needs. If you only need to trade in the S&P 500 in the US, there are, in our view, perhaps a dozen firms that would have a realistic chance of achieving best of breed designation for you. However, if you need to trade in US and Europe or trade a lot of mid and small caps, ADRs or ETFs, the field narrows. If you add in Asia or South America, you are down to a handful at most. The optimum way for you to choose best of breed providers is to stack them up against each other and look at each of these criteria. Engage in real market conditions with each, then take a step back and ask, What value am I getting across these different measures? This is the best way you can gain a clear understanding of best of breed. n 35