Automated Trading and the New Markets

By Robin Sharp
May 27, 2006

Automated (or Algorithmic) trading refers to computer systems that trade financial instruments without any human intervention. New automated trading systems are changing both the face and the infrastructure within investment banks. Vendors and banks are now exposing more of their systems to programmatic trading opportunities as their e-commerce offerings mature.

Over the next ten years, these changes will continue to evolve and create radical changes in the investment banks. This article explains the software reality of the new trading systems and the profound effects they will have on the banking infrastructure.

History

Automated trading has been around for well over 15 years in the equities market. In that time open outcry trading floors have gradually been replaced by automated exchanges. The original and still major share of automated trading is in the equity markets. Equity trading was amenable to automated trading because the market was centred on the exchanges. Automated equity trading now accounts for 20-25% of all trading in these markets. Because the FX, Money and Bond markets trade OTC (Over the Counter) they are not centralised (i.e. exchange based), it has meant waiting longer for automated trading to be enabled.

Last year saw the close of the last open-outcry trading floor in the City of London – (the International Petroleum Exchange) – which has since re-camped, on smaller scale, in Dublin. The previous ten years have seen the gradual replacement of manual-exchanges (open-outcry) floors with software exchanges in the equity, futures and options markets. Over past 15 years over the counter trading in the FX, Money and Bond Markets has also moved from phone to on-line trading. The in-house trading floors in banks have, until recently, been relatively unaffected by the drift from phone trading – as traders have simply moved from phone trading to on-screen trading. Automated trading has been restricted to the equity markets with other trading done thru phone calls and vendors on-screen trading systems.

In the past few years vendors have begun to expose their software interfaces for instruments (e.g. FX, Money and Bond Markets) that were previously hidden behind these on-line trading screens. This has enabled the potential for blanket market coverage of automated trading across most of the “primary” markets. Financial institutions are now facing the challenge that other financial institutions will gain a competitive advantage by exploiting these new trading opportunities to their disadvantage. The problem for these institutions is that it is not clear how these new automated markets will pan out.

As I will explain it would be naïve to assume that the development of automated trading in these new automated markets will go the same way as the equity markets. The naive view would be to assume that 20-25% if these markets will become automated and trading will carry on as normal. I hope you will understand by the end of the article why these new automated markets will have a far greater impact than the rise of automated equity trading.

Vendors

The origins of automated trading started with the Vendors who offered prices quotations from other traders on the exchange floors and other trading floors. Originally these prices were indicative (and not tradable) but under competitive pressure vendors were allowed to publish tradable prices. Once tradable prices were available in the vendors quotation screens, vendors extended their service to allow traders to execute those prices. Finally vendors enabled automated trading by exposing their software.

In the past couple of years vendors, such as Reuters, EBS and Bloomburg are now offering automated trading across all the underlying instruments (e.g. equities, FX, bonds and money markets).

It is worth remembering that some trading is still done entirely by hand, such as trading of structured derivatives. In this market customers must still ring around banks for the best price because the deals are complex.

Banks

Another origin of automated trading has taken place within the banks own infra-structure, and started in the dot-com boom. Originally a few banks offered non-tradable prices on their websites to a few private clients, in a few instruments. Since then need to compete has meant more banks have been forced into offering more tradable prices, in more instruments at faster timescales and small price increments. Unlike vendors (e.g. Reuters) banks have not been able to offer on-line screen trading and have had to offer automated trading via web interfaces.

As this increasingly competitive market of automated trading has increased the banks have been forced to become increasingly hands-off as humans stop being able to cope with the volumes and speeds. Typically a banks system trading environment is now multi-tiered and receives and offers prices through multiple feeds. Depending on the market and size of the bank traders will face electronic systems from Exchanges, Brokers, OTC, Clients and other Banks.

The diagram below illustrates a tiny snapshot of a banks electronic trading environment:

Messages

A further development has come from some of the message oriented organisations, such as FiX (equities) and FpML (FX). These organisations had originally provided standards for the exchange of messages between back-offices. In past few years they have extended the scope of their messages to include quotation and execution of instruments. The standardization of quotation and execution messages will means a quickening of the pace of development as more banks must now compete on a level playing field.

Progress

The history of automated trading can now be traced fairly clearly, and has involved the gradual removal of hands from the trading process.

(1) Hands-on Quotes - Prices are quoted over the phone-to-phone
(2) Indicative Prices - Prices are published but require manual confirmation
(3) Screen-based Trading - Prices can be executed on the screen
(4) Automated Trading – Prices can be published and executed by computer

Most importantly, the evolution of increased automation between counterparties (banks, vendors and clients) has now come to an end. The process of automation could become quicker and slicker, with more counterparties and more instruments, but no new degrees of automation can be added between counterparties. Automated trading can now be enabled across all the primary instruments, and it is anticipated that a lot of the derivative instruments will become automated too. The question is therefore, within banks, where will automated trading go next?

Now that the automation between counterparties has been achieved this will enable banks to start to focus on automation within the bank. Given the structure of banks current trading systems automated trading can be predicted to process in the following way: -

(5) Automated Trade Netting - Netting for orders in the market
(6) Automated Vertical Trading – Instrument are broken down to their vertical components
(7) Automated Horizontal Trading – Instruments are priced across traditional markets
(8) Hands-off Management – as strategies are executed by computer

However, as history never repeats itself it’s highly unlikely that such radical changes to trading systems will result in a copy-cat computer versions of the current software reality, where traders are simply replaced by machines.

Let’s go back to the current situation and look at the development of this sequence starting with automated trading data to see how automated trading could develop within the banks.

Automated Trading Data

The diagram below is a typical example of trading data. The bid represents what people are willing to buy the instrument at. The offer represents what people are willing to sell the instrument at. The diagram shows the best bid and the best offer. There may be other bids and offers in the market, but these are not shown because they are not told to you. The distance between the bid and offer is called the ‘width’ of the market.

For every bid and offer there may be one or more traders contributing that price. Behind these traders there may be many more traders offering slightly worse bids and offers (sitting ‘outside’ the market). The number of traders bidding and offering at any one time is called the ‘depth’ in the market.

When the price people are willing to buy equals the price people are willing to sell, the bid equals the offer, the price is set, and a deal is done. If there are 3 traders are bidding and 5 traders offering at the same price then only 3 deals are done. It’s a first come first served basis, so the traders that put their price in first will be first to be dealt. In the FX markets for example all this typically takes place between 1 and 3 seconds.

There are a few things to notice. The offer is always greater or equal to the bid. Conversely the bid is always less or equal to the offer. The distance between the bid and offer is called the width of the market. You can always submit a bid equal to the offer, or offer equal to the bid. Some automated trading systems stop you from putting a bid above the offer.

There is what I call a ‘lemon’ between each deal. I call them lemons because the shape of the width is typically one of divergence and convergence. A price inside the lemon would be referred to as ‘in-the-market’, or simply ‘on-market’. A price outside the lemon would be referred to as ‘out-the-market’, or ‘off-market’.

Price Feed

When you subscribe to a vendor for price data on an instrument, the information you get varies slightly from vendor to vendor, but in general you get the following information:

• Market Bid
• Market Offer
• Market Bid Depth
• Market Offer Depth

The Market (or best) price data includes data that has not been credit screened. This means that you will not be able to deal on that particular price because you do not have a credit arrangement with the counter party.

• Tradable Bid
• Tradable Offer
• Tradable Bid Depth
• Tradable Bid Offer

The Tradable (or creditable) price data includes data that has been credit screened. This means that you can deal with the counterparty. In some cases the Market price may be better (or worse) than the Tradable price, but you can’t trade it.

• Local Bid
• Local Offer
• Local Bid Depth
• Local Bid Offer

The Local Bid price data includes prices that have been offered by traders within your own bank. If you are about to submit a price into the market and find that a local trader has already done so you may wish to trade locally with them, rather than go out into the market, because it means you will not pay brokerage fees. This is not a risk free operation, as the local traders price may get dealt before they have chance to cancel their order in the market and deal with you, and you will then have missed your chance to trade at that price.

• Last Dealt Bid
• Last Dealt Offer

Some vendors will give you the last dealt price in the market. In a ‘traditional’ market feed the bid will generally be below the offer price, however if that is true all the time, no deals would get done. So watching this value is important because it gives you closure on the lemon. Knowing this value is useful to predict the behaviour of the market.

Micro-Economics

The simplest way to understand this price matching behaviour is to go back to the basics of laws of demand and supply. The laws form the basis of micro-economics. The number of bids on the bid queues represents the number of traders who are willing to buy the instrument at the bid price. The number of offers on the offer queue represents the number of traders who are willing to sell the instrument at that price. If we look at the classic demand and supply curve below we can see that the higher the price the less people wish to buy and the more people wish to sell, conversely the lower the price the more people want to buy and the less people want to sell.

However, there are problems with this classic view of economics, when trying to determine the next tradable price of an instrument.

The biggest problem is that you only know about the highest bid and lowest offer, not the whole graph. Even when you are told about the number of traders on the bid and offer queue, this is limited to up to maximum number of say 8 or 20. You do not know how many traders are waiting outside the market on different queues at the different prices. This means you can’t predict the slope of the graph.

A related problem is for the majority of the traders the instruments value is relative to the price other traders are willing to pay for it. The same trader within a few seconds can go from bidding or offering one price to one that is very different which is still is close to the market price. Price is relative, and changes on a second by second basis.

It would be extremely difficult to go out and conduct a cross-sectional study of the market to find out what the various traders are thinking at any moment in time. For a start you couldn’t conduct the survey quick enough, and nobody would tell you what they are willing to pay even if you could ask them fast enough. Similarly it’s just as difficult to conduct a longitudinal study, of individual traders, as you don’t know who has submitted which price you cannot build a profile up of an individual trader to determine if they are bearish or bullish.

Whilst it is possible to gain some leverage using micro-economics, building up a curve is quite difficult. A more realistic (but complex) approach needs to be found. As we will see, the reason this is important is that if some understanding of economics at this "quantum" level can be obtained then it will undermine traditional trading and banks will be forced to compete at this level.

Now we're going to go a little further into the mechanics of automated trading so we can realise why the recent automation across the primary instruments could mark a qualitative shift in trading practice from the existing situation where only equities are heavily traded automatically.

Submitting and Cancelling an Order

When you submit an order into an automated trading system you must submit the following information.

• Instrument
• Price
• Quantity
• Bid or Offer

The instruments could refer to equities foreign exchange, bonds or forward interest rates. The price would be the price quoted in the market specific convention. The amount would be in a market convention, such as a minimum amount, or units – such as units of $1 million. Finally you need to say whether you are bidding or offering the instrument. For example you could bid $5 million for the FX instrument GBP/USD at 1.6789. This would mean you were offering to buy $5 million dollars for £2.9781 (5/1.6789) pounds.

Similarly you can cancel an order that you had placed. For example you may wish to cancel an order where you were bidding (to buy) at a price you think may be too expensive (and you would get ripped off), or bidding (to buy) at a price you think may be too cheap (and you would never get hit). Conversely, you may wish to cancel an order where you were offering (to sell) at a price you think may be too cheap (and you would get ripped off), or offering (to sell) at a price you think may be too expensive (and you would never get hit).

The Bid-Offer Queue

Looking at the bid-offer matching process in a little more detail we see that for any moment in time there are a number of bids and offers in the market. These bids and offers sit in matching queues. There are a number of queues for bids and offers at different prices. In the first diagram below you can see bids and offers in the market, but the price does not match. In the next diagram the bid price has changed and there are now 3 bids and 5 offers at the same price. Each bid in turn at the matched price will get matched. In the example below, the first 3 bids and offers will get matched.

If you some reason you cancel an order then re-submit it at the same price you will go the back of the queue and risk not getting dealt. In the example above we can see an offer [1] being cancelled, then re-submitted and going to the back of the queue. This could be caused by a timeout or other market event. In this case the offer [1] has gone to the back of the queue and won’t get traded.

Price Ticks

We're now going to give you a look into the timing of price ticks. I'm telling you about this mostly to add more complexity to an already complex picture and to help you to understand the speed at which automated trading works. It’s important to understand the frequency that the data that arrives down the wire. The underlying timings of the price ticks can have an important effect on the interpretation of the data. These effects include which prices you receive and how stale those prices are. The important issues with price ticks is the timing of the distribution of price tick data and that it is deemed fair for all the traders who are subscribing. There are basically 2 different models that vendors use to supply prices:

1. Snapshot Pricing

In the snapshot pricing model the vendor sends out a snapshot of the price of an instrument every second, on the second. This means that when your time slice comes up you will be sent a snapshot of the market at that point in time. Each trader is given a different time slice, so they all receive the data at different points in time.

In this model it is important that the incoming data is not triggered by any event other than a timing event. This is so each trader is sent a snapshot of the market with an effective random delay between the quote arriving and being sent to the subscriber. This is deemed fair.

The disadvantage of this method is that the price staleness may be quite large. However in a busy market this effect is minimal.

2. Delta Pricing

In the delta pricing model the vendor sends out the price to every trader in a round-robin manner, with up to one price a second. The round-robin is constantly rotated so that if you received the price first you will receive it last the next time a price is sent.

The disadvantage of this method is that is takes longer to get all the prices out to all the traders when a price arrives because the processing is more intense. This method also has the disadvantage that the delivery of price information is loaded towards the front of the second, so if the price changes in the second half of the second you will not get told about it.

Trading Process

There is a saying in the UK that a week is a long time in politics. Well, a second is a long time in trading, and in one minute you can lose your days profits. Whilst the theory behind program trading is fairly simple, the software reality is that program trading operates at a different time-scale to even the fastest human trading.

Click image for a larger version

The diagram above shows the timescales of a typical human-human trade-life cycle. The business process that is going on here is that the trader on the left has a think then submits a price into the market. Subsequently the trader on the right accepts that price and the trader on the left will be notified.

An important thing to notice is that the human effort of perception and reaction make up the largest percentage of the life-cycle’s time-frame. When we look at the total time frame (excluding the traders initial and final thoughts) we get the following table:

This may seem like an exaggeration but it is well known in the psychological literature that it takes 0.5 seconds for a human to become conscious of a visual stimulus. Looking at (or even being) a trader gives you an impression that trades are being done quickly, as a reaction to a price on the screen, but as an observer you lose 0.5 second becoming conscious of the price changing, and 0.5 seconds becoming aware of the traders reaction. As a consequence the reaction appears almost immediate, but it isn’t.

The human aspect of trading could be reduced further. For example if a trader had a stop-loss or take-profit order where they were simply waiting for a price and reacting then these timing figures comes down to about 500 ms human effort. However, those sorts of "dumb" decisions can be made by computers, and we're interested in how everyday trading can be undermined by automated trading because the timescales that computers can operate at are faster than human reaction times.

Timescales

In computer trading the timescales come down drastically from human trading. In human trading decision making is done on the basis of the speed that the brain can process incoming information and the nature of the information. In some cases, such as fund management, information is assimilated based on years of experience. In other cases, such as spot trading the information is close to the market data and decisions are made within a few seconds.

It’s worth understanding that the brain brings experience and subjectivity to the table and software brings speed and objectivity to the table. For most types of trading experience beats speed but there is a lot of noise in the market in the sub-second region where the brain simply can’t compete with a computer.

Diagram shows the time scale of the trade duration vs. the frequency of information type.

Click image for a larger version

The diagram above illustrates the relationship between information input and time, and the type of trading that is done. As time-scales get shorter there comes a point where the conscious human element of trading is left behind and a new market involving subconscious trading happens. At the moment the volume of trading at these sub-second time scales is not be great (< 5%) and is held back because the coarse granularity of ticks and price data has been designed for human interaction. However this will change. In the past year, banks and vendors competing with each other for customers have started to offer tick data and price amount at finer resolutions and the scope for trading below the conscious level will now increase.

There are two basic types of trading that can be done below the levels of human consciousness. These are:

• Psych-Trading
• Systems-Trading

Psych Trading

Possibly the most important strategy in any contest is to play to your opponents weakness. Psych-trading is about understanding the weakness of human trading and trading against those weaknesses. The human mind is a remarkable instrument but at short time-scales, fast changes, complex calculations and high volumes it’s weaknesses provide opportunities. Additionally most people don’t realise there are also opportunities in psych-trading that operate over longer periods of time.

This face and vase ‘ambiguous’ picture illustrates psych-trading. In classic automated trading computers represent one face and the trader is the other. By contrast in psych-trading the computer is not a face but the vase. In psych-trading the computer sits between the traders and exploits their weaknesses.

Psychologists have comparatively recently been studying how the conscious model of the human subject is built up over a 500 millisecond time period, and how new information is assimilated into this conscious model. If you understand the weaknesses of the human mind you can understand how the mind can be disrupted by incoming data and react to make a profit. Depending on the nature of the changes the conscious model can be completely destroyed, interrupted or certain changes can be missed. Similarly information processing bottlenecks in conscious processing mean that certain changes in market prices may not be consciously available, or may take longer to consciously process by traders. Similarly understanding the limits of mental processing is important. A large number of trades are difficult for traders to juggle in their heads. When such large events occur is small time frames computers can predict irrational behaviour of traders, again for a profit.

A different style of psych trading is to try to model the market sentiment as a sort of macro-economic, or macro-psych modelling, rather than the micro-psych model just mentioned. Macro-psych modelling, in the form of sentiment, operates at different time-scales. The formal definition of sentiment is an emotion about a subject (for example pride in a flag). By understanding the dynamics of sentiment human subjectivity (not their experience) can be exploited. For example over long time scales human sentiment can over inflate prices, in the classic bubble. Understanding the detailed psychology helps you to understand the clues. There are some stage magicians who use these techniques to great effect. Knowing these various ‘irrational’ human traits can help when deciding when to go into or cancel trades in the market even in the very short term.

System Trading

In contrast to psych-trading, system trading is all about build computer models of the vendors computer systems. So far we have looked at various queuing and ticking aspects of some vendors systems. Knowing the intimate differences in vendors systems is very important when making refined decisions about trading in the market. When trying to build these computer models it’s important to throw away the human side of the market, because from a computers perspective the vendors systems are the market.

Once you have built a simple model of vendors systems you will need to refine it. Trying to do this manually is quite difficult. The timing between vendors systems are often below human perception so the best strategy is to monitor different aspects of vendors systems on a real-time basis. For example it’s necessary to understand how fast it takes to be matched on a trade because there is a counter party on the queue, because this will tell you when you could be in for a long wait.

Another example is if you know that vendor A’s price is 100-300 milliseconds behind vendor B then you can use this information to do longitudinal arbitrage trading. These decisions can get even more refined based on specific instruments. For example if vendor A publishes its Dollar New York Prices through a slow gateway to London then you can pass those prices directly to London to get ahead of the market. Vendors systems are not static either. You may find that given certain market conditions or times of day trading will be more profitable. Building a system that dynamically models vendors performance becomes a key factor for success.

It is important to design system trading computer software to provide data mining and graphic visualisation of system data so that analysts can spot trends and patterns in the data. Once trends and patterns are spotted the system should be dynamic enough to allow rapid configuration and deployment into the live trading environment with minimal disruption. The market reality is that once a profitable pattern is spotted it may not last for very long, so the software reality is to spot these trends quickly and work in a very agile deployment environments.

Trading Volumes

Submitting an automated trade is not free, there is a cost – usually between $5 and $20. This means that when you submit an order into the market the probability of making money needs to sufficient to cover your brokerage costs. An alternative strategy is not to go into the open market at all, but to try to trade internally, within the bank, before going to the open market. Trading within the bank not only reduced brokerage fees but also keeps assets within the bank. As trade automation becomes more prevalent within banks the opportunity to trade internally will increase.

Trading internally involves making two changes to automated trading software. First a matching-engine needs to be built that contains a bid offer queue for all the instruments. Second each automated trading system needs to try to trade with the internal matching engine before it trades with the external market. This is no small task. Political decisions also need to be made, for example who keeps the profit made by a trade matching engine or should it trade its own positions.

Traders are already finding it increasingly difficult to monitor the trades, when the automated trading systems are first put in traders cope, but as volumes increase they cannot monitor the volume of trades flowing through the front office. Trading will move from trade management to position management and the role of trade monitoring will become a role in its own right.

Vertical and Horizontal Automated Trading

With the opening up of new markets to automated trading new opportunities will be found. In the current trading environment traders typically trade a small subset of instruments within a market. For example you may get a trader who trades FX in one currency against the Euro or Dollar, or a trader who trades Forwards in one currency. Some of this delegation is due to trader specialisation, but the bigger factor is that traditional corporate management structures and regulatory structures impede cross-market trading. With the introduction of trade automation cross-market trading is more easily implemented. Cross-market trading can be done either vertically or horizontally.

Vertical cross-market trading refers to trading instruments and their derivative or underlying instruments in a managed way. Computers can manage ‘vertical’ automated trading in two ways. They can monitor the prices of the slower derivative markets and exploit them by splitting derivatives into their component parts and trading them at a profit – vertical arbitrage. Computers can also construct derivatives on demand and publish quotes, competing with existing human markets by bringing the spread down to near-market levels – vertical construction. Because derivative markets are more closely aligned to their underlying instruments in the human-management and regulatory frameworks trading across these markets would be expected to be done first.

In this simple example a Forward FX instrument is broken down into its Forward and FX components (vertical), and a Bond is broken down to its FX and UK equity components (Horizontal).

Just to repeat the point made earlier, in the short-term vertical cross-market trading will be more prevalent because the corporate management structures are more closely aligned to this type of trading. In the longer-term ‘horizontal’ cross-market trading will appear. Potentially this form of trading will be more profitable because there is more liquidity in underlying cross-market instruments than in any market for an instruments derivative. Risk management systems already understand how to price instruments in terms of instrument in other markets. For example equities, forwards and corporate bond markets are closely related. One of the reasons traders don’t trade cross market is that they cannot price the instruments fast enough, and – again – is that traditional corporate management structures and regulatory structures impede cross-market trading

Once the pricing of more complex instruments becomes automated then this will increase the opportunity for cross-market arbitrage and will mean traditionally slow moving desks will be forced to improve their spreads and response times in order to compete.

The diagram below illustrates the complex relationships between instruments that are currently handled by option and structured traders in a less timely manner than their FX or equity counterparts.

Click image for a larger version

Eventually arbitrage will force separate markets to revaluate their relationships. It only takes one successful arbitrage engine to forever link two previously unrelated markets. Anybody in the business will know how deeply this will be felt.

New Software for New Trading Markets

I have no doubt that in 10 years time the trading environments will be a lot tougher, and that they will have got tougher quicker than people have expected. Pressure from well resourced and agile companies, such as the hedge funds will put pressure on banks margins.

Automated trading currently accounts for 20% of trading in the equity markets and 5% of trading in FX markets. In future there will be increased volume in the fast-end of the market. Once this software is in place there will be a slow growth in automated trading into the slower markets to handle hedging, position keeping, position building and daily trade management.

I don’t believe there will be any real “artificial intelligence” in the new trading markets. Neither is it going to be down to brute force speed. As I have said the network latencies of c.100 milliseconds will have an effect on the need for performance, give or take a millisecond. The success criteria of the new trading software will be down to understanding the psychology, systems and instrument relationships rather than intelligence or speed.

The new trading software will need to have a very different shape to existing app servers with their over designed pattern-oriented software. The new trading software will be more demanding to engineer because it will be both real-time and cross-instrument. Manual trading and pricing will assume less of role on the desktop in favour of dashboard software, where the traders will configure and “baby-sit” the automated-trading systems. The line between traders and software engineers will blur and traders will have to become more like programmers and programmers more like traders. This will create questions for the regulators about who exactly is doing the trading and who authorised the trading models when they are very flexible and dynamic.

Monitoring trading systems profitability will become increasingly important. Losing money on a trade has a probabilistic component so it’s difficult to tell whether why you lost money without going into the trade in detail. There is a need for banks to have detailed analytic software in place that will determine which aspect of the trade that caused the loss. This trade analytic software will have the following requirements:

1. Collect the Data on a millisecond by millisecond basis

2. Visualise the relationships between trades and market data

3. Statistically analyse the different aspects of then trade

Clever use of graphs can help program traders and analysts to spot trends. After seeing enough “post-mortem” graphs the viewer can generate and test hypotheses much quicker than by building software models.

The diagram below shows a very simplistic representation of a trade. The blue and red lines depict bid and offer respectively. The orange lines show the depth in the market. The green lines indicate trades being submitted or executed.

In the example, after an offer for 3M is put into the market and is hit for 2M then 1M a few seconds later. Clearly there is more depth in the offer than the bid when the offer goes into the market. Then the depth of the bid improves and the 3M trade is hit for 2M. Subsequently the dept on the bid improves again, indicating that the market price will rise. The computer decides to stay in the market and is hit for a further 1M. Whilst the trade executed successfully it may have been a better trading strategy to pull out of the market once the depth in the bid improved because it was indicating that the price was about to rise. Making this decision involves incredible sensitivity to the market in terms of both psychology and systems.

Summary

Banks are beginning to change their trading infrastructure to deal with the new automated markets. This process will continue to develop over the next 10 years. New software will have to be developed by any front office wishing to compete in these markets. Over the next 10 years increases in cross-market trading will put pressure on different divisions within the bank to co-operate, and this could be the most difficult aspect of automated trading to achieve.

Robin Sharp has worked in the financial markets for 20 years and has a degree in “Psychology with Computer Models” from the University of Sussex.

<< Back to Soapbox

<< Back to the Front Page