The Gateway to Algorithmic and Automated Trading

AT Round Table: Multi Asset Algo

Published in Automated Trader Magazine Issue 03 October 2006

Multi asset execution algorithms have been touted as the next “big thing”. AT asked four major sellside providers for their viewpoint. With: - Tim Wildenberg, Head of Direct Execution Services Europe, UBS - Jeff Wecker, CEO of Townsend Analytics, a subsidiary of Lehman Brothers - Kevin Bourne, Global Head of Execution Services, HSBC - Jarrod Yuster, Head of Global Portfolio and Automated Trading, Merrill Lynch

How much buyside demand do you feel there really is for multi asset algo (MAA) trading? Who do you see as the most enthusiastic buyside segment?

Wildenberg: While we are having some conversations with clients about MAA trading, I certainly wouldn't say that demand was overwhelming. However, there is certainly some expansion of interest in trading new asset classes individually using algorithms. For example, we are seeing more clients enquiring about using algorithms for trading futures, but that interest still isn't exactly overwhelming. I think that is partly because futures markets are relatively liquid and cheap to transact in, so the requirement to come up with a better and less expensive algorithmic means of trading is not as strong as in equities.
Before MAA trading starts to grow significantly, some changes will first need to take place in multi asset (nonalgorithmic) electronic trading. In that environment, there are relatively few clients who connect to us in more than two asset classes over the same connection.
Many of them are still happy to use a separate set of tools per asset class because they are still themselves to some extent structured along asset class lines. (The person trading equities is not also trading fixed income or FX.)
I would say derivative-based or hedge fund type clients, as opposed to institutional traders, were showing the greatest interest in MAA trading at present.
Wecker: There are investment strategies where algorithm demand is higher than others and in these cases algorithms for the efficient placement of multi asset class trades are already in use. A simple example of this would be equity index arbitrage where algorithms are used to finesse the placement of the individual stock and index futures trades.
Index arbitrage is a fairly standard and long-established type of trade, and while, formerly, I don't think there were a lot of algorithmic offerings in this space, some buy and sell side firms have now integrated the algorithmic placement of multiple asset classes into their systems. Another example of active MAA trading is, of course, the simultaneous trading of an overseas stock and its associated FX hedge.
I would say the most enthusiastic buyside segments would be quantitative institutional asset managers and hedge funds. Quant managers often base their models on the relationships among securities, and it would not be unusual for them to have a model for the relationship across asset classes. In some cases, those money managers are constrained to operate in a single asset class and they may just use the signal from other asset classes to guide their trading. In other cases, a quant manager might actually have a mandate to trade multiple asset classes.
For the hedge fund manager, trading constraints tend to be much more relaxed. I think that most of the growth in demand for multi asset class algorithms will come from the hedge funds, and, of course, from sell side proprietary trading.
Bourne: I think that while there may be a certain amount of smoke and mirrors in evidence at present, in the long term there is a very real intent. When the technology enables the capability, then there are already more than enough market participants who understand where the potential alpha is in multi asset algorithmic trading.

Tim Wildenberg: "Before MAA trading starts to grow significantly, some changes will first need to take place in multi asset (non-algorithmic) electronic trading."

Tim Wildenberg, Head of Direct Execution Services Europe, UBS

Initially, interest in MAA has been coming overwhelmingly from the hedge funds, but with many conventional asset management companies now owning their own hedge funds, the interest is broader than it appears.
Yuster: There is obviously a lot of demand already for algorithmic trading in cash equities, but we are also seeing real demand in equity options and futures. Clients now have technology on their own desktop that gives them considerable flexibility as to whether they route orders to brokers, crossing networks, or exchanges. As the client's desktop becomes more pervasive across different asset classes, clients will want more sophisticated algorithms for trading across those classes.
I would say MAA trading is driven by clients who are sophisticated in terms of technology and quantitative skills. That certainly includes hedge funds, but you also have the very sophisticated pension and institutional funds as well. They have their own technology organisation and their own quant group, but haven't always received much recognition or credit for their influence in driving electronic trading using equity algorithms. Finally, you can also see changes on the sellside that presage growth in MAA trading; people moving across asset classes and/or taking responsibility for multiple classes. The intention is to try to create more synergies, cross selling opportunities and technology efficiencies.

How much buyside interest just relates to the ability to trade multiple asset classes through one interface and how much to true MAA trading in the sense of synthetichopper.jpg instruments etc?

Wildenberg: We are starting to see people talking to us about gaining access to multiple products through the same interface, which makes economic and logistic sense. But that is driving the requirement to do multi asset electronic trading, rather than multi asset algorithmic trading.
There are some people who are trying to create synthetic structures or to have a broader, more complete trading perspective on a particular instrument they are looking to trade. In the latter case, they may only want to trade a company's bonds, but also want visibility on its credit derivatives, equity and commercial paper. However, even these clients are not demanding an algorithmic package that covers all those bases.

The demand to support multi-asset class trading in the front-end will likely outstrip the demand for algorithms that integrate the trading of those asset classes. The reasons for this are not necessarily directly related to the algorithms themselves. Technologists and operations managers at a buy side firm will typically find it far simpler to buy one product to support the trading of multiple asset classes rather than different products for each asset class. At that firm, there may indeed be a few individuals who need multi asset class algorithms, but there are plenty of other drivers why the firm would use a single platform to trade multiple asset classes, such as support, cost, and simplicity of external relationships.
Bourne: I think the answer is both. People want to do multi asset class trading through a single FIX interface because it is operationally efficient. Many investment banks are less keen on this idea because they have historical infrastructure that is split across business lines and is thus not well-adapted for this purpose. Clients will not want to pay a separate fee for an individual trading interface per business line, nor will they want to attempt multi asset class trading in such a fragmented manner.
As regards synthetic instruments, we have received a lot of requests for synthetics in equities globally - especially from hedge funds. To date, I don't think there has been a lot of activity in trading pre-constructed synthetic instruments using algorithms. However, there is definitely buyside interest in this area and the general growth of automation in (for example) equities will make it a practical possibility.
Yuster: At present, I think it is more the former in terms of gaining efficiency improvements by being able to see and risk manage multiple asset classes through one interface. I think we'll see that extending across more asset classes over time in terms of view and risk management. Ultimately that will extend to more sophistication in terms of multi asset algorithms.
I think the first phase is to have algorithms for the individual asset classes, where algorithms will take in data from a range of instruments and sources. For example, equity algorithms might take into account the stock being traded, the industry, and any associated derivatives, such as stock index futures. Initially, such algorithms will only be trading the equity, but they will be sourcing and using data from other assets. The logical progression from there is to actually trade these additional assets as part of a broader strategy across classes. It is a similar evolution to that which took place in equity algorithms when they developed from just single stock algorithms to portfolio algorithms.
In other words, the first step towards MAA trading will simply consist of trading additional instruments using their individual algorithms. The next step is to produce true cross asset algorithms, which is a much more sophisticated process.

How much of the interest do you think is in finessing the placement of multi asset trades and how much in just automation?Whats_Next.jpg

Wildenberg: The demand I see is more based around the desire to automate efficiently. Once that has been accomplished once for one product/asset, it makes sense to extend it across others. The logical extension of that is to add algorithms on top as well, but that is typically the second step after automation is already in place.
Wecker: I believe that the interest is more about finessing the execution rather than automation. When you say automation, I assume that you mean simple slicing of component orders without any modelling of the returns process for the fluctuation of the underlying securities. For an MAA algorithm to work effectively, it would have to make certain assumptions about the market and the relationship among the different assets. Before a trader is going to trust an algorithm to execute components of her strategy, she needs to be convinced that there is value to be gained from that algorithm. This value ultimately has to come in the reduction of the slippage costs necessary to complete the trade. Slippage cost can be measured in a lot of different ways, such as comparing the actual execution prices to those prevailing at the inception of the strategy. It is not clear to me that automation alone without more sophisticated order placement will provide the advantages necessary to off-set the risks of using an algorithm vs. manual execution.
Bourne: I think it depends who the client is - for some it is a case of both. It is not just about transmission of the order in terms of operational efficiency, but also ongoing maintenance of the portfolio. If you have a synthetic portfolio, your costs of maintaining it can be substantially reduced.
Historically the hedge funds have been big drivers of this. As they have become larger they have realised that operational infrastructure has a very direct impact on the returns of the fund. Because they have a propensity to trade using more complex trading structures they feel the impact of complex operational cost quite acutely. By nature their transactional model is three-dimensional; therefore their operational model is three-dimensional.
Yuster: I think it is a combination of both, but there are limitations. You want to automate in order to allow the human trader to operate at optimum efficiency, but you obviously don't want to automate just for the sake of automation. There's no point in automating trades where the computer cannot outperform (or at least equal) the human trader.

What do you think will be the order of asset/instrument adoption in MAA trading? Is FX inevitably next?

Wildenberg: I would say algorithmic trading is definitely flavour of the month as regards FX. The central limit order book price time priority model of the equity market is starting to percolate into foreign exchange markets, such as Hotspot. This is effectively encouraging the sort of algorithmic thinking we have seen in equities.
However, in terms of actual activity I think that at present more is happening in the futures space, which is a logical extension of the cash markets because the linkages are often the same. The same technology can provide access to both markets, for example on the Tokyo Stock Exchange both cash equities and TOPIX index futures are available on the same exchange platform.
As yet I see little interest in or discussion of fixed-income.
The key point that underlies the order of algorithmic adoption is the amount of infrastructure already in place. For example, a primary reason that equities saw such huge growth in algorithmic trading was that the bulk of the client base was already connected electronically to its brokers. The underlying groundwork and foundations were already there.
However, while that is also true in FX and futures, it is rather less certain that the necessary back-end plumbing into order management and portfolio management systems has been done as it has (because of the FIX protocol) in equities.

Jeff Wecker: " For an MAA algorithm to work effectively, it would have to make certain assumptions about the market and the relationship among the different assets."

Jeff Wecker, CEO of Townsend Analytics, a subsidiary of Lehman Brothers
Wecker: I think foreign exchange and futures are running neck and neck, with a lot of opportunities in both.
Bourne: I think FX is happening already. I think you're going to see on-demand FX prices coming from the investment banks via FIX. These will start to be fed on a large scale into the algorithmic models of the hedge funds. In fact this is already happening today with some of the larger hedge funds.
The demand for FX as part of MAA is also evident from the increasing number of requests for FIX based FX prices we have been receiving from long fund order management system (OMS) vendors. This demand has been created by the OMS vendors' clients wanting to execute both their overseas equities trades and the related currency hedges simultaneously through one interface. Rather than doing the FX hedge later with their custodian (probably at an inferior price) they are able to deal with the currency risk instantaneously.
Yuster: I think foreign exchange will almost certainly be the next asset class to take off. By contrast, fixed-income is a much more difficult market in terms of the way it trades and as regards data capture. However, I think clients will start to incorporate it into the MAA mix over time if the right tools are available.

Do you see any significant obstacles to MAA trading? How might these obstacles be addressed?

Wildenberg: There are to my mind two fundamental issues:
The first is the challenge of accommodating the different structure of various markets, which may require very different types of algorithms. For example, equity algorithms in the US have been primarily concerned with aggregating liquidity across multiple trading venues. By contrast, algorithms for European equity markets are more focused upon trading intelligently into a single point of liquidity against a benchmark to improve upon execution quality. Extrapolate that to other asset classes and you would probably place FX and fixed income (which have multiple points of liquidity) in the same category as US equity markets. By contrast, futures would be more akin to European equities. This means before considering the minutiae of algorithms for an asset, one has to take into account these structural differences - you cannot simply point an existing algorithm at a new asset class and expect it to work.
The second issue is the need to have the necessary operational structure in place. Executing multi asset trades is one thing, processing them in the back office is quite another. A related challenge is managing customer credit across multiple assets, which is easy to talk about but difficult to do. For example, if a client trades equity against FX do they receive one contract note that summarises the two trades or do they have two trading relationships? If you provide one contract note, you cannot then allow the client to net across other trades they may do with you. This sort of issue tends to be glossed over, but is actually hard to resolve in practice.
Wecker: First, a substantial research effort is necessary to build successful algorithms. I would consider that effort, complex as it is in a single asset class, to be more complex than just the sum of the parts when you consider multiple asset classes. It is not just the sum of the complexities of the individual instruments, because there is an interaction component to consider as well. Understanding and modelling the covariance or correlation among assets would be important in building a successful MAA algorithm. For many "equity" algorithm teams, covariance modelling is new research with new issues to consider.
For these teams, there is no free lunch - they have to do the work. I think you may find in these early days that a number of players will start by promoting relatively naive algorithms that just slice orders up for each security in a multi-asset classes trading strategy. The industry will then quickly move well beyond that, and algorithms will take into account ever more sophisticated models of the interaction or relationship among the different asset classes and the securities within them that underlie the strategy.
The evolution of algorithmic trading in equities illustrates this point. Early on, a lot of firms claimed to have similar VWAP algorithms, but there were key differences among VWAP algorithms. There were those that just implemented time slicing and there were those that actually looked at the process of how the underlying stock fluctuated in relation to shifts in liquidity. I think you'll have a similar evolution in the multi asset class world, with one possible difference. Those who have already developed sophisticated algorithms in single asset classes might have an edge, because the incremental amount of work that they need to do will be less than for those who are starting from scratch.
Bourne: I think for a lot of banks the obstacles are historic and related to the traditional product silo mentality. They have built their infrastructure based upon their old organisational model, where each business unit had its own technology. This obviously doesn't fit well with clients wishing to trade seamlessly across asset classes.
Another problem for some banks is maintaining a single credit view of clients across all areas of interaction. In a multi asset class trading relationship, the bank might be providing clients with OTC prices on some instruments and taking market orders from them on others. The challenge is that these two situations are significantly different in terms of risk profile. For example, committing a quote on a fixed income instrument represents a commitment of capital, while taking a market order represents delivery versus payment (DVP) risk. Coping with this sort of situation requires a real time multi asset class risk capability, which represents a major investment that not many banks have made.
Kevin Bourne, Global Head of Execution Services, HSBC

Kevin Bourne: " The demand for FX as part of MAA is also evident from the increasing number of requests for FIX based FX prices..."

A related issue is that if a bank is taking a lot of trades from clients who tend towards synthetic risk across asset classes it has to be able to calculate how it will make the necessary market plays to hedge the exposure. To do that effectively requires a common data model across all asset classes, and again this is something that by no means all banks have. Without it, one cannot price multi asset trades effectively or understand the impact of that pricing for the bank across the various markets.
Yuster: I think that there are a number of obstacles that the sellside has to address if it is to facilitate the evolution of MAA trading. In the first instance, there is a core set of crucial technologies that require a considerable effort to implement in a scalable fashion. These will allow the efficient handling of large transaction volumes at high frequency. While the deployment of such technologies require appreciable investment and effort, if accomplished successfully in a sufficiently generic manner then they can be redeployed for multiple asset classes.
Another crucial area is data infrastructure. Every possible relevant data element for every asset in every asset class that will be going into the MAA melting pot needs to be gathered. Unfortunately, gathering the data may only be the first step. In many cases the raw data may be less than clean and algorithms built using such raw data are unlikely to prove robust. Therefore a considerable investment in terms of data cleansing will be required, but such a cleaning process must not "over-cleanse" by removing valuable information along with noise. The ultimate objective is to have the most comprehensive possible picture of the events that occurred in the market - including the order types that are specific to each market.
This process also needs to recognise the various markets' cost structure rules. For example, markets such as NASDAQ have very high trade frequency but one isn't penalised for high trading frequency because charges are based per share rather than per trade. By contrast, certain markets in Europe and Asia charge per trade, so it is therefore essential to model them on the basis of a lower trading frequency.
Apart from the need for a robust quantitative data infrastructure supported by high quality quantitative research, MAA trading requires a single credit relationship. At present there are very few banks who can handle multiple asset class trading with clients on the basis of a single consolidated credit relationship.

Where do you see the greatest competitive differentiator for the sellside in MAA trading for hedge funds? (E.g. back office integration or algorithm design)

Wildenberg: Having the ability and willingness to make the necessary investment. Providing the necessary algorithms and platform for just a single asset class is extremely expensive. The investment required for an electronic trading platform is in itself substantial, but overlaying that with the technology required for algorithmic trading adds another level of complexity and cost.
For example in equities you might have an algorithmic trading engine processing ten thousand trades a day, but each of those trades will have many fills, and each order might be broken into many sub orders. Therefore you simply cannot afford to have a failure in the technology that supports this, because the chances of recovery in such a complex environment are negligible. In addition, there is the need to recruit the very highest calibre quants to design and build the algorithms.
If you then multiply the costs of all this across multiple asset classes and add a further premium to allow for any inter-asset complications, you begin to gain an inkling of the scale of the investment required in order to be competitive.
Wecker: Before embarking upon using an algorithm for MAA trading, a multi asset class trader is probably already selecting his providers on the basis of how well coordinated they are across the asset class product silos they offer. A provider that has achieved a high degree of coordination across asset classes may have already worked out how to provide clients with efficient financial leverage across multiple asset classes as well. The capital required to finance a multi-asset class position is an important factor for the trader to consider.
If a provider has passed all the tests necessary to support a trader who is trading in multiple asset classes without algorithms, that provider will probably enjoy a competitive edge if the investor wishes to start using algorithms to trade. I believe that these are the providers who are most likely to invest in developing multi asset class algorithms, because they already appreciate the value created by being able to deliver execution services across multiple asset classes.
There will be a number of factors that differentiate sell side firms offering multi-asset algorithms. The firm that can conclusively prove that their algorithm reduces the slippage of a trade across asset classes will have a clear advantage. In addition to providing the algorithm, a firm needs to prove that they have excellent electronic connectivity to all the various asset classes being traded. Also, the firm must have a suitable, and ideally, easy to use front-end to deploy the MAA algorithms.
Having the right service model around the algorithm is also crucial. An investor cannot use an algorithm they cannot understand. Therefore the quality of the provider's training and support for their algorithm will play a large part in determining whether or not it is widely accepted. In this respect, it is important to realise that the people who design and build algorithms are not the same as those who service and support them. A firm that is embarking on the challenge of delivering multi asset class algorithms as part of their execution service suite therefore needs to realise that there is a complete business model that has to be built around the initial research investment in order to succeed.
Bourne: I think it is risk - a single client risk profile that runs across all asset classes. For sophisticated buyside clients who might want to trade across twenty or more diverse assets at the same time on an automated basis this is an absolute requirement. In our experience, one of the first questions such clients ask is, "Can you provide a single credit relationship across asset classes?"
Yuster: I think it is a combination of algorithm design and ease of use and efficiency. A lot of clients talk about transaction cost analysis but it is hard for them to create a uniform measurement for this across brokers that they can then normalise. I typically see clients choosing brokers now for algorithmic electronic trading based upon relationship, ease of use, but also operations. If you are going to devise complex algorithms that will straddle asset classes you have to be able to support these operationally in the back office.

Do you feel that the buyside has really thought through the trade synchronisation implications of MAA trading across multiple venues? (E.g. legging risk)stepper_croc.jpg

Wildenberg: Once robust tools are available for MAA trading, they will be able to manage that synchronisation. If they don't, then one won't have achieved MAA trading in the true sense.
Those on the buyside capable of building their own multi asset algorithms will by definition have the necessary quantitative and technology skills and will undoubtedly have thought through this sort of issue.
Wecker: I think the percentage of the buyside that both understands the implications and knows how to construct a solution for dealing with them in a multi asset class environment is very low. For the average trader, it is relatively easy to understand the implications of trade synchronisation among asset classes or underlying securities that comprise a strategy.
However, having access to an easy and automated solution to minimise slippage in the most productive way is not readily available to as many of these traders. The spread between the number of people that understand these trading issues and the people that know how to program computers to solve them creates the opportunity for algorithm development.
I also believe that some proportion of the multi asset class strategies that are winning today are a result of exploiting intelligent automation. One result of increased multi-asset class algorithm development will be a likely reduction in the returns from some of these strategies.
Bourne: While the more sophisticated members of the buyside are well aware of such implications, I think overall there is still a lot of value lost through inadequate pre-trade transaction analysis, including legging risk. In some cases, clients don't even properly factor in the various market charges (commissions, spreads etc) of all the trade legs.
However, you can see that some clients clearly do appreciate the trade synchronisation implications of MAA trading across multiple venues by observing the way in which average transaction sizes for multi asset trades have been declining. Those trading these more complex strategies, such as hedge funds, are having to reduce the individual trade size in order to ensure that they can complete all the legs of the trade in a timely manner without incurring legging risk.
However, this has implications in terms of operational costs - for example, a higher bandwidth requirement to accommodate the greater transaction volume count.
Yuster: I think the clients who are pushing multi asset algorithmic trading are the most sophisticated clients, who are very well aware of the implications. However, the average buyside participant isn't. You can see this in just the equities space when some of them want to trade equities across multiple regions - unfortunately one has to accept that it isn't feasible to trade Japan, Europe and the US simultaneously.
Jarrod Yuster, Head of Global Portfolio and Automated Trading, Merrill Lynch

Jarrod Yuster: " If you are going to devise complex algorithms that will straddle asset classes you have to be able to support these operationally in the back office."

Do you think the way some exchanges/markets are broadening their asset/instrument bases will materially assist MAA trading or actually hinder it by causing further fragmentation?

Wildenberg: The technology that has evolved to deal with the challenge of US equity markets can already deal with this sort of situation. So while you can argue that fragmentation is an issue, the technology is already available to cope with it, so it is no longer a practical concern.
Wecker: Both are likely to happen. Exchanges that broaden the array of asset classes that are available in their trading venue will make it easier to develop multiasset class algorithms provided, of course, that they are able to attract sufficient liquidity across all the asset classes. However, as more venues compete for liquidity in the same asset class then the increased fragmentation that may arise in a particular asset class will require increasingly more sophisticated algorithms that aggregate the trading of an asset class across multiple trading venues. That will increase the complexity of the algorithms and make them more difficult to create.
Bourne: I think exchanges are still a long way away from providing access on a single order book to enough instrument classes with sufficient liquidity/variables so that those such as hedge funds would be able to identify cross asset class trading opportunities within them. I don't think there is currently enough depth of information outside the equity type asset classes for this to work.
Now that very large global tick databases are available from those such as Reuters this permits hedge funds to do new types of modelling and to research new opportunities in a way that was never possible before. I therefore don't think they have as yet exhausted the possibilities within equities alone.
Having said that, there is no doubt that the exchanges are extremely interested in this space. We have been approached by two major exchanges (that have also approached other banks) about putting our algorithmic controllers inside their firewall.
The logical (and perfect) extension of this (beyond running a highly liquid multi asset class order book) is that you would have a futures exchange operating in the same location as a cash exchange with both using a shared data centre. Banks would then be able to put their algorithmic platforms inside that data centre where they would be fed by tick data coming from both exchanges in realtime. That gives you the fastest possible solution, the maximum information and the greatest number of trading opportunities across asset class.
In a technical sense you could argue that this model represents the next level in algorithmic trading. Up until now the focus has been on optimising performance between the broker and exchange gateways to reduce latency. If algorithmic platforms are located inside the exchange, then this competitive space disappears and the only thing left to optimise is individual application performance. At the same time, this will render worthless the substantial investment many banks and brokers have made in building the lowest possible latency direct market access (DMA) platforms.
Yuster: If the sellside provider has smart order routing and algorithms that take advantage of this then the possible fragmentation won't make much difference. After all, sourcing disparate and hidden liquidity has become a standard discipline in the algorithmic trading of equities. Expanding this to accommodate additional asset classes and trading locales obviously requires additional work, but if the basic generic infrastructure is already in place then this isn't an insuperable problem.Equity__FI_legs.jpg

Many of the markets now being talked of for MAA trading (e.g. fixed income) are radically different in terms of trading characteristics from equities. How do you think execution algorithms for those markets will differ from those already used in equities and how will they combine with equity algorithms for MAA trading?

Wildenberg: As I mentioned earlier, when it comes to designing algorithms there are two broad types of market to consider. The first is fragmented (like US equities) and is essentially a series of quote driven OTC trading venues, which require algorithms capable of finding and aggregating liquidity. The second resembles European equities, where the emphasis is more on algorithms that can improve on price. FX and fixed income would fall into the first category, futures the second.
Wecker: The process used for the trading or auction of a security dramatically affects the algorithm creation process. It is much easier to construct algorithms where you have executable bids and offers in the asset classes you are trying to trade. Once you get into a marketplace that is request for quote (RFQ) driven, rather than streaming quotes, the development of an algorithm can become far more challenging.
Also, some RFQ marketplaces have rules in place to limit access from computer-generated trading, be that algorithms or black box trading. An algorithm designer would have to overcome the execution challenges of that asset class well in advance of considering that asset class as part of a multi asset class execution algorithm. I wouldn't rule it out, but I would say some markets need to see more evolution, before their participants target them for multi asset class algorithms.
Bourne: At present I think the cross asset opportunities are still rather limited. It is possible to do MAA trading on some of the electronic crossing networks that have fixed-income instruments, but most of the liquid instruments here are government, not corporate, bonds. There are some limited opportunities in trading government bonds versus equities, but I think that until peer-to-peer relationships through FIX become prevalent for fixed-income, the bulk of the buyside will have limited interest in MAA trading. This is because they will be unable to understand in real time the relationship between fixed income quotes and equities/derivative market tick data.
A key factor here is and remains data. I see multi asset algorithmic trading being very much driven by the buyside's ability to collect prices and create the historical databases against which they can model and make trading decisions. In equities/listed derivatives this is already available and it is just starting to happen with FX. Once that spreads across other asset classes then things will start to take off. However, one hurdle here is that in decentralised markets sellside firms will not share their historical data with each other, which will make it very difficult for them to build certain cross asset algorithmic trading models for the time being. Increased post trade transparency from legislation, such as MiFID, may start to change this if applied across all asset classes.
Yuster: I think fixed-income will evolve in much the same way as foreign exchange has in terms of clients obtaining direct market access. I think this evolution has to take place before any meaningful MAA progress can take place.