The Gateway to Algorithmic and Automated Trading

Algorithms Where do we (and you) go from here?

Published in Automated Trader Magazine Issue 06 July 2007

A whole industry has developed around algorithmic trading. Connferences, exhibitions, publications carrying articles (not dissimilar to this one) and advertising are but a few of the activities that are providing employment and profits for many. On top of this,one must consider the costs to sell-side providers of employing project and programme managers,developers, programmers, connectivity specialists,integrators,systems, trading and support teams.Not forgetting,of course,sales people and marketing budgets needed to make the potential client base aware that the broker or bank in question is in the 'game'.

Jon Carp

Jon Carp, Head of Alternative Execution Sales Europe at CA Cheuvreux

While this list is far from exhaustive, it serves to make the point that the business of providing and continually developing algorithms that match and ideally outperform competitors is a very expensive and time consuming one. The cost means that only a select few of the sell-side community can develop and provide their own algorithms for client use. But having attained clients, will the rewards be profitable in the longer term?

The answer, of course, is not straightforward. At CA Cheuvreux we have developed and continue to develop a suite of algorithms that perform well. These are used internally by our own agency traders and programme desk, and have been developed further to satisfy client requirements. However, algorithms should not be viewed in isolation, but as part of a complete service offering. Experience has shown that very few, if any, of the buy-side use algorithms as their only method of trading. Direct market access and programmes continue to rise in usage. Orders that require sell-side traders to utilise their specialist knowledge of local markets have not, despite what some will lead us to believe, gone away. To maximise returns on developments in trading techniques, a complete product set should be offered to clients. In many cases, we have won a client because we have the entire suite. Feedback from our client base has shown that although some only require a singular service, others assert that to only offer DMA is insufficient, likewise for algorithms, programme trading and telephone execution. By offering all services, with no gaps, client attainment and retention is far greater than it would otherwise be

By no means is CA Cheuvreux the only company to offer a multitude of trade execution services. Indeed, providers of algorithms generally offer a suite of the same eight or so basic strategies (i.e. VWAP, TWAP, % Volume etc.), with perhaps one or more developed to distinguish their offering. But if the basic strategies are the same, then there is almost too much choice for the user. We have found that it is common for buy-side traders to use, on average, three algo sell-side providers with upwards of five providers not uncommon. This is maybe due to the increased use of algos to pay for research services, unbundling and as the use of electronic trading methodologies continues to gain favour. So, if a fund has five providers each providing ten general strategies, it has a selection of 50 general strategies to chose from, each with their nuances. In addition, it has the individual providers' 'specialist' algorithms to choose from as well.

That is a lot of choice. Basic economics tells us that return on investment will not be sufficient in a situation of oversupply. Indeed the effects are already being seen. Some sell-side firms are already opening up their systems and algorithms to enable competitors to connect their clients, at a cost. My guess is that this trend is driven by providers looking to squeeze some extra return as much as to satisfy client requests for multi- broker systems.
Taking all of the above into account, where can we see the future of algorithm development and distribution?

With the multitude of options now available, many to essentially do the same thing, a trend seems to be developing. Software houses, including order and execution management system vendors, have started to react to client demand and opportunity. They are now producing algorithms on behalf of the buy and sell side. A broker can now 'buy-in' a suite of basic algorithms and distribute them to clients at a relatively low cost. Suppliers range from large companies to specialist algorithmic boutiques, many of the staff having worked in the banks and brokers, integral in algorithmic product developments. Likewise, the buy-side can either write their own or purchase algorithms from vendors that perform in a satisfactory way. As vendors begin to provide more sophisticated models, this means that the buy-side will have access to a core suite of algorithms that fit, or are tailored to, their own individual needs - independent of sell-side influence.

The natural conclusion of this trend is that the brokers will in future receive order flow on a DMA basis, but would not be able to charge a premium for providing the algorithms. Meanwhile, the buy-side would re-coup their investment via the reduced commission rate.

So where does this leave the sell-side? The first, seemingly obvious, conclusion we can draw is that if third parties supply adequate products, the sell side will be able to cut costs on algo developments, or at least reallocate resources. In reality, brokers will continue to develop real value-added algorithms that go further than the basic solutions available today and in advance of what third parties are able to offer. To continue to attract business, brokers will develop and publicise products that show a real unique selling point. This has already started with banks devising strategies, with highly creative names, that take algometric trading further into a complex world.

At CA Cheuvreux, we recognise the market trend and, whilst the usage of our algorithmic suite increases month by month, we also acknowledge the need to continually gather client feedback and develop with close reference to this. We are fully aware that the cost and resource required to be all things to all men is large and that products that satisfy requirements are readily available 'off the shelf'. We look to utilise what is best of breed and for the situation. After all, why develop and recreate something that already exists, works and is available? Surely it is better to spend time developing new and innovative strategies? We have not yet taken up the offer from some of our competitors to act as a sales channel, i.e. to use, distribute and pay them for their algos, as we believe our own provide well for our clients. But never say never.

With Mifid fast approaching, the creation, development and launch of new liquidity pools requires the necessary investment in systematic solutions that satisfy new regulation. This surely gives just one example where innovative development of new algorithmic strategies can reap rewards for those that can see the potential opportunities.

In conclusion, the algorithmic industry has been on long a journey of development. That journey, however, still seems be at stage one with following stages likely to be equally long, competitive and, therefore, fascinating.

cheuvreux