GB2366023A

GB2366023A - Credit limit storage in an anonymous trading system

Info

Publication number: GB2366023A
Application number: GB0101435A
Authority: GB
Inventors: Neena Jain; Steven Iaccheo; Edward Howorka; Vladimir Neyman; James Shu; Michael Merold; Raj Krishnasami; Gregory Mills; Alistair Gladwin Crane; Roy Steven Mcpherson; Paul M Ginsberg
Original assignee: EBS Nominees Ltd
Current assignee: EBS Nominees Ltd
Priority date: 2000-06-23
Filing date: 2001-01-19
Publication date: 2002-02-27
Anticipated expiration: 2021-01-19
Also published as: WO2002001437A2; US20020133455A1; CA2383125A1; AU7663401A; GB2366023B; JP2004502243A; ZA200202217B; US7693774B2; EP1295234A1; GB0101435D0

Abstract

An anonymous trading system comprises a network of broker nodes each of which perform price matching, deal execution and market distribution. Trader terminals on trading floors are connected to a given broker node via a trading agent node. Credit may be stored for a given trading floor at its trading agent node or at a credit agent node which stores credit limits for a group of trading floors. Where the credit limits of a party to a proposed deal are stored at a credit agent node, the deal execution process is routed through that credit agent node to check for available credit. The system allows institutions to adopt any combination of local and global credit assignments with counterparties with which it wishes to trade.

Description

2366023 CREDIT LIMIT STORAGE IN AN ANONYMOUS TRADING SYSTEM

FIELD OF THE INVENTION

This invention relates to electronic brokerage systems and in particular to systems in which counterparties trade 5 anonymously within fixed credit limits. Such systems may trade financial instruments such as foreign exchange and forward rate agreements. The invention is particularly concerned with the handling of credit limits between counterparties in such systems.

10 BACKGROUND TO THE INVENTION

A number of anonymous trading systems are known in the art. EP-A-0,399, 850, EP-A-0,406,026 and EP-A-0,411,748 all assigned to Reuters Ltd disclose aspects of an automated matching system in which a host computer maintains a central 15 database of bids and offers submitted by terminals connected to the host via a network. The host also maintains records of credit limits between each trading bank and the possible counterparties with which it is willing to trade. The host computer uses information in its central database to match 20 bids and offers and buy and sell orders based on matching criteria which include the counter party credit limits.

Generally, counterparty credit limits are set for each bank or each trading floor and the host computer establishes a gross counter party credit limit for each possible pair of 25 counterparties. The gross counter party credit limit is the minimum amount of remaining credit between two counterparties.

A trader's terminal will display a subset of the trading book, typically the best few bids and offers. These will be updated periodically to ensure that the trader sees the true state of the market.

A problem with the system outlined above is that the trader sees the bids and offers irrespective of whether he 5 has sufficient credit with the counter party submitting that bid or offer to trade. As a result, a trader can attempt to trade when there is no available credit. As the system is anonymous the trader has no knowledge of the counterparty until a trade as been completed and so, when he hits a bid 10 or offer, has no idea as to whether it is likely to be accepted or rejected for lack of credit. This is extremely frustrating for a trader, particularly in a fast moving market in which trading opportunities can easily be lost.

The problem arises as the host computer only checks 15 available credit after a deal has been proposed and a potential match identified.

This problem was solved in W093/1S467 now assigned to EBS Dealing Resources inc. Instead of displaying the actual trading book, or a part of it, to each trader, a different 20 market view is shown to each trader in which bids and offers from counterparties which whom they have insufficient or no credit are screened out. Thus, the trader only sees prices with which he knows he can deal.

The architecture of the system of W093/15467 is very 25 different from the of the Reuters system and is based on a distributed network with a number of arbitrators which perform matching. Actual credit limits are stored at local bank nodes to which each of a bank's trading terminals are connected ensuring that sensitive credit data does not leave 30 the bank's physical site. The actual trading book is sent by the arbitrators to the market distributor. The market distributor forms a market view specific to a given trading floor and sends it to the relevant bank node. A different market view may be formed for each trading floor depending on credit criteria. Thus, the market view which is distributed to each of the bank nodes is the complete market view with credit screening taking place, the market distributor to filter out any prices with which the bank, or S a given trading floor within the bank, has insufficient credit.

In addition, the market distributers also have limited credit information, maintaining a credit matrix which may store a simple "yes-no" credit indicator for given 10 counterparties. When a match is made, the prices having already been screened for credit, the bank node will make a second credit check using the credit matrix to see whether any previously extended credit has already been exhausted.

While both the above systems have been used 15 successfully in the financial trading markets for a number of years, they both suffer from the disadvantage that they are very inflexible in the way they handle credit. They require banks to tie up large amounts of credit in one area of their trading activities. A typical bank will be trading 20 a number of financial instruments and a number of different markets and will want to trade up to its credit limits in each trading day. If one particular market is quiet it will want to be able to divert the credit assigned to that market to a different field. Similarly, if a particular market is

25 very active it will want to be able to take advantage of that activity. It should be remembered that a given bank may be dealing with many of the same counterparties in different markets. It is thus undesirable to tie up credit to trades in one particular instrument with a given 30 counterparty as this may diminish the bank's trading capacity within its own global trading limits. Also, the manner in which credit limits are assigned does not reflect the manner in which many banks handle and assign credit. This results in a System which has to be treated differently, as far as credit is concerned, by a bank which discourages banks from using the system.

SUMMARY OF THE INVENTION

The present invention aims to provide a system in which 5 the manner in which credit is assigned is more flexible than the prior art systems.

An embodiment of the invention provides an anonymous trading system for trading financial or other instruments which include a communications network for transmitting electronic 10 messages. A plurality of trader terminals are connected to the network, each can generate electronic price quotation messages which includes bid and/or offer prices and can communicate to the trader information from other trader terminals. One or more matching engines are connected to is the network for matching bids and offers and, when a match is made, for executing deals. A market distribution means connected to the network distributes price messages to the trader terminals and is responsive to price quotation messages and the matching engines. Credit limit storage 20 means store credit limits available for trades between a trading floor or group of trading floors and possible counterparty trading floors or groups of trading floors. The credit limit storage means includes at least one credit agent node which stores credit limits for a group of trading floors.

The use of a credit agent node enables credit for an institution to be handled globally. That is, credit does not have to be allocated on a floor by floor basis. This allows users of the System the flexibility of assigning credit from 30 trading floor to counterparty trading floor, from one of a number of groups of trading floors to individual or a number of groups of trading floors, from institution to institution or any combination of these local and global credit arrangements. This allows users of the system greater flexibility than prior art systems and assist in the full utilisation of available credit limits during the course of 5 a trading day.

Preferably, the trader terminals are connected to the network through trading agent nodes. The credit agent node may also be a trading agent node for simplicity. Each trading agent node which does not hold its own credit 10 information for connected trader terminals includes means for sending credit enquiry messages to the credit agent node on receipt of a proposed deal message. The credit agent includes a means for receiving the credit enquiry Messages for the trading agent nodes, for checking the available is credit and for indicating to the system whether the deal can proceed. Preferably, the indication may show that deal can proceed for the part of the proposed amount if there is insufficient credit for the whole deal. Preferably, the credit agent node is not directly connected to the trading 20 agent node and credit enquiry messages are sent via one of the matching engines. It is preferred that the deal indicator message from the credit agent is sent to one of the matching agents which routes it to the counterparty trading agent.

25 In one preferred embodiment, the credit limit storage means of a given party interfaces to an external credit limit storage means of that plurality. This enables credit within the system to be treated as part of that party's overall credit scheme for that instrument, which may be 30 traded through a variety of media, or that counteifparty with which it may trade a variety of instruments.

The system may comprise many credit agent nodes.

In one preferred embodiment of the invention, an automated trading system for anonymous trading of instruments, including financial instruments, comprises a computer communications network having a plurality of 5 interconnected broking nodes and a plurality of trader agent nodes to each of which are connected to one or more trader terminals comprising a trading floor. The trading agent nodes are each connected to one of the broking nodes and each broking node comprises a matching engine for matching io bids and offers input into the system as price quotation messages from trader terminals, means for executing deals where prices are matched, and market distribution means for distributing price messages to the trader terminals in response to price quotation messages and the matching is engine. one or more credit agent nodes are provided each of which store credit limits for trading floors connected to a number of trading agent nodes for trades with Possible counterparty trading floors or groups of trading floors.

Those trading agents whose connected trading floor credit 20 limits are not stored at credit agent nodes include a credit storage means for storing their own credit limits.

BRIEF DESCRIPTION OF THE DRAWINGS

An embodiment of the invention will now be described, by way of example only, and with reference to the 25 accompanying drawings, in which:

Figure I is an overview of a trading system embodying the invention; Figure 2 shows the flow of messages when a new quote is submitted in the system; 30 Figure 3 depicts the production of a market view to traders; Figure 4 shows the flow of messages when a trader submits a buy or sell order; Figure 5 shows the flow of messages to update broker nodes following a buy or sell order; Figure 6 showa the flow of messages when a broker modifies a quote; 5 Figure 7 shows the deal execution process; and Figure 8 shows the flow of messages in a global credit model; DESCRIPTION OF PREFERRED EMBODIMENT

10 The present invention will be described with reference to the dealing architecture illustrated in figures 1 to 6 and which will be hereinafter described. However, it should be understood that the invention is not limited to that architecture but could be implemented in any anonymous 15 trading system. For example, it could be implemented on either of the Reuters and EBS Dealing Resources prior art systems known in the art and referred to earlier.

The electronic brokerage system to be described provides a platform for trading at least the following 20 instruments: FX (Foreign Exchange) Spot, FRA's, and Forwards and also FX Forwards, CFDs, short-dated government and/or central bank paper, commercial bills, CDs, inter-bank deposits, commercial paper, repos, interest-rate futures, swaps, options and various tailor-made variants on these 25 basic products. These are all referred to as financial instruments. It may also be used for trading non-financial products such as commodities.

Traders at trader terminals are connected to a communications network which allows electronic messages to 30 be passed between terminals. These messages include the submission of quotes and hits which are then passed on to each of a plurality of broker nodes throughout the system. A quote is a bid or offer order submitted by a trader to \make a market" and is distributed to other traders as part of a market view. Quotes are thus orders visible to other traders. A hit is a buy or sell order submitted by a trader wishing to create a deal on the basis of a price displayed 5 on his market view derived from one or more quotes. Hits are orders which are invisible to other traders.

The computer trading system of Figure 1 comprises a plurality of trading agents 10 each connected to at least one of a plurality of broker nodes 12. Each trading agent 10 is the means by which the trader terminals access the trading system with a given trader terminal being attached to one or more trading agents.

Trader terminals (not shown) may be workstations or other computer terminals configured to generate and submit 15 electronic order messages including bid and/or offer prices, quotes and orders (usually through use of a specialised key pad) and to communicate market view data, including price and amount available, for financial instruments to be traded. The communication is usually by display but could 20 also be by printing the information, voice synthesis or otherwise. The trader terminals may input orders manually or automatically. In the latter case a trader may program a terminal to submit an order once the market reaches a predetermined state. Alternatively, the trader terminals may 25 bot be individual work stations as such but integrated into an institutions automated dealing capability. All these variants are examples of order input devices.

Traders are typically grouped as part of a financial institution, such as a bank, which arranges traders as part 30 of a trading floor. A trading floor is a group of traders under common control of a trading floor administrator who allocates credit lines for the trading floor against other trading floors. The market view for a trader, or group of traders, is the market information (price, volume, etc.) that the traders can see that reflect th e market. The market views are preferably pre- screened for credit compatibility, as described in W093/15467 the contents of 5 which are incorporated herein by reference. Thus, traders only see displayed quotes with which they can trade. As well as extending credit to a trading floor, credit may be extended to a bank as a whole (many banks have several trading floors indifferent locations), or to an individual 10 trader or group of traders. Credit may be extended from the bank as a whole, from a trading floor or from an individual trader. This process will be described in detail later.

The system is an anonymous trading System in which the market views produced by the brokers comprise price and 15 amount information without identifying the source of the price. The prices displayed for available bids and offers and the amounts available at those prices, are thus aggregates of one or more quotes. Only the quotes of parties satisfying the pre-screen credit criteria are 20 included in the aggregate price displayed. The market views produced by the broker nodes thus differ from one trading floor to another depending on the credit allocation.

The trading agent node provides services to a specific trading floor or group of traders. These services include 25 providing access to the network for each trading work station, completing deals, producing deal tickets and maintaining historical dealing information for traders. Each trading agent node must connect to at least one broker node to access the trading system. A group of trader 30 terminals thus connects to a trading agent 10 to 'access the system.

Each Broker node 12 provides the basic order matching and price distribution services. The Broker nodes are arranged in a structure called a Clique Tree which enables faster communications routing, following very specific but simple rules. The Clique Tree is a network structure where individual nodes are grouped into Cliques, and the Cliques are then arranged into a tree structure. Each Broker can be 5 linked logically to a number of Brokers, which are referred to as its neighbor Brokers. Communication between Brokers is on an equal level, with no "up" or "down" direction in the network.

In the embodiment of figure 1, there are three Cliques:

10 that formed by brokers 12a, 12b and 12c, that formed by brokers 12b, 12d, 12e and 12f and that formed by brokers 12e and 12f. It will be seen that brokers 12b and 12e are both in two Cliques. It will be appreciated that, for ease of understanding, only a simple example is illustrated. In 15 practice the network may be very much more complex with a large number of cliques.

While Trading Agents must be connected to at least one Broker node, they are not members of the Clique Tree, but remain outside the structure. A Trading Agent connected to 20 multiple Broker nodes will receive multiple sets of market prices. Even though the price information from different Broker nodes can be substantially the same, the information may be received at different intervals. A Trading Agent will send a given trading order to only one Broker node.

25 The term Broker node is used to describe a computer arranged as a physical or logical node in a computer network providing a broking function. The basic broking function is the storing of quotes, providing the quotes to traders in the form of a market view and matching quotes and orders.

30 The Broker nodes in the described embodiment also perform further functions, but these are not essential features of what is defined as a Broker node.

Thus, the broker nodes each provide a matching engine which is connected to the network for matching submitted bids and offers and, when a match is made, for executing deals. They also perform the function of market 5 distributors distributing prices messages to the trader terminals in response to the price quotation messages and the matching engine. Within the context of the present invention it is preferred that the matching and market distribution functions are amalgamated in the broking node 10 but the invention is equally applicable to systems in which the functions are separate and performed at geographically and/or logically separate locations. An example of such a system is W093/15467 referred to earlier.

The Broker nodes are equal to each other, and perform is the same functions. The arrangement of the network or their position in it is transparent to the broker nodes. They only need to know about their neighbours. Each Broker node has knowledge of all orders in the market, and is able to match orders as soon as they are submitted. As each Broker 20 node maintains a full list of orders in the market, it is therefore able to customize market views as needed by the Trading Agents and is able to react faster to market information as soon as it is received.

To understand the purpose of the distributed broker 25 node arrangement, price distribution and deal execution will now be described with reference to figure 2.

The deal process begins with one or more traders submitting orders into trader terminals. An order is a dealing request from a trader, with instructions to buy or 30 sell with specific restrictions, such as price and amount. A quote is a persistent order that remains available in the system and is distributed as part of the market price information. Quotes are used to "make the market", and are known to traders as bids or offers. A hit is an order that has "invisible" and "fill or kill" properties ("invisible").

Hits are not distributed as part of the market price. A hit does not remain in the System; if it can not be dealt when 5 entered, it is removed.

An Order Book is a list of all the available orders in the market. Since the Quotes are the only available orders, the book consists of a list of Quotes. The Quotes are arranged in a queue in the correct dealing order. The sort 10 order of the queue may vary for different trading instruments. The default sort order is by price and time. In the system, each Broker node maintains a complete list of all available quotes. In a system such as foreign exchange there will-, effectively, be two books, one showing orders to buy and the other showing orders to sell.

The message flow in the system is described by named messages, each carrying appropriate parameters throughout the network. The process of submitting a quote (persistent order) begins when a Trading Agent receives information from 20 a trader workstation that a trader has issued a bid or offer. The Trading Agent then starts the quote submission process. When the Trading Agent receives the quote information from the trader workstation, it will create and maintain a context for the quote. It will then send a Quote 25 Submit message to the Broker node that it is connected to. The Broker node will validate the quote and accept it if valid. This first Broker node that receives the quote becomes the "owner" Broker node for this quote. In example shown in Figure 2 this is Broker node 5. This is the only 30 Broker node that can commit the quote to a deal. The Broker node will create a context or "quote object" and sort it into its queue for the correct tradable instrument.

After the quote is placed into its queue, the owner Broker node will then distribute the quote throughout the network by sending QuoteAvailable messages to other Broker nodes. In this example, Broker node 5 sends the QuoteAvailable message to Broker nodes 2 and 6. As each Broker node receives the message, it creates a context 5 (quote object) and sorts it into its queue (order book). It notes in the context which Broker node had sent it the message. After placing it into the queue, the Broker node then sends the QuoteAvailable message on, using broadcast routing rules, to all neighbours except those in the same 10 clique as the broker who sent the message. Therefore, Broker node 2 sends it to 1, 3 and 4. Broker node 4 then sends it to Broker node 7. At this point, all Broker nodes know about the quote, and update their order books accordingly.

The broadcast routing rules are applied to ensure that 15 network traffic is handled in an efficient manner and to reduce any duplication of message flow.

The broadcast rules are:

1. The Broker node originating information will send it to all of its neighbour Broker nodes.

20 2. A Broker node receiving the information will send it to all of its neighbours Broker nodes except those in the same clique as the Broker node that sent the information.

3. If a message contains persistent information, 25 such as a quote, the information will be stored with the identifier of the Broker node from which the information was received.

Note that these rules refer to the information, not the message that contains it. For example, information about a 30 quote may be sent to one Broker node in a ProposeDeal message and to another Broker node in a MarketUpdate message. However, the same information is sent to both Broker nodes, and so the above rules apply.

Price distribution is the process of providing market information to the traders at the trader terminals. This 5 information is created by the Broker nodes and sent to the Trading Agents for distribution to the traders. This process is shown in Figure 3.

Each Broker node will examine its queue of quotes (order book) and calculate a view of the market for each 10 Trading Agent connected to it. This view is built specifically for the trading floor that the agent represents. Views may be different based on credit or other factor3. The exact process for determining a market view will vary based on the trading instrument. The view 15 information is sent to the Trading Agent in a MarketView message. It follows, therefore, that each of the brokers hold credit information for each trader and the possible counterparties.

Hitting a quote is the basic process Of creating a deal 20 between two traders. A hit from one trader is matched to a quote from another trader. This process is shown in the figure 4. The Trading Agent of the trader terminal hitting a price shown on his market view display sends a HitSubmit message to the Broker node. This message targets a price, 25 not a specific quote. The Broker node will scan its queue and find the first Quote in the queue that can be matched with the hit. The matching rules may vary based on the trading instrument.

When the hit is matched to a quote, the Broker node 30 will modify its context for the quote, moving the amount matched from "available" to "reserved pending deal". This will prevent the same amount of the quote to be matched with another hit. The Broker node will then send a ProposeDeal - is - message to the Broker node from which it received the quote.

This message will target the specific quote. In this example, the hit comes from a trader connected to a trading agent connected to broker 7. Broker 7 will send the message 5 to Broker 4.

As each Broker node receives the ProposeDeal message, it checks the quote in its queue. If the amount of the proposed deal is still available in the queue, the Broker node performs a similar process as the matching Broker node.

10 The amount of the proposed deal is moved from "available" to "reserved pending deal". The ProposeDeal message is then sent to the Broker node from which it received the quote. In the example, Broker node 4 sends it to Broker node 2. Broker node 2 will then send it to Broker node 5.

is The routing of a ProposeDeal message follows targeted routing rules. Targeted routing is used to deliver information to a specific Broker node. Since knowledge of specific Broker nodes is not built into the system, the target is not a specific Broker node, but is the Broker node 20 from which the information originated. For example, a message is not sent to "Broker node 714", but is sent as to "the Broker node originating quote 42". The targeted rules are:

1. A Broker node originating a message about a specific piece of information will send the message to the Broker node from which it received the original information.

2. A Broker node receiving a message about a specific piece of information that it did not originate, will send the message to the Broker node from which it received the original information.

The message will thus follow the path of the original information back to its source. In the example this is from Broker node 7, to Broker node 5, via Broker nodes 4 and 2.

When the Broker node that originally created the quote receives the ProposeDeal message, it performs the same checks and amount reservation as the other brokers. Since this Broker node owns the quote, it has the authority to commit the quote to a deal. The ProposeDeal message represents the authority to commit the hit to the deal. The 10 Broker node will then initiate the deal process by sending a HitAmount message to the Trading Agent that submitted the quote. The deal execution process is described later.

As the deal matching process takes place, it is necessary that the list of quotes maintained at each Broker is node be keep up to date. This is accomplished by each Broker node notifying others when it makes a change to a quote, as shown in figure 5.

As each Broker node changes a quote in its queue, it notifies all neighbour Broker nodes except those in the 20 clique from which it received the change. In the example above, Broker node 4 received notice of a change in a quote from Broker node 7 in a ProposeDeal message. It notifies Broker node 2 by sending the ProposeDeai message. Broker node 4 must now notify Broker nodes 1 and 3. This is done by 25 sending a MarketUpdate message to these Broker nodes.

Following the normal routing rules, the information about the quote is distributed to each Broker node in the network. Any Broker node receiving the MarketUpdate message will pass it to all neighbours not in the clique from which 30 it is received. Note that a Broker node sending a ProposeDeal message should not also send a MarketUpdate message to the same Broker node. This would result in duplicate information beingreceived and the deal amount being reserved twice.

When the deal matching process is completed, as described above, the deal execution process begins. This 5 process completes the deal and commits the traders to a deal. The process is shown in Figure 6. As matches are made and deals initiated, information is made available for traders. This information can be used to inform a trader that a deal is pending. Any given trading application can 10 decide if the trader should be informed. In any case, the information is available.

The Taker's Trading Agent will be notified as soon as the initial match is made and the ProposeDeal message is sent. This agent can notify the traders workstation at this is time. This pending deal information may change as the matching process continues. The maker workstation is notified of the pending deal when the maker's Trading Agent checks credit and sends the DealStatusMaker message.

The deal execution process begins when the maker's 20 Trading Agent receives a HitAmount message from its Broker node. This message informs the Agent that a match was made for one of its quotes. The message identifies the quote as well as the amount of the hit, counterparty and the identity of the hit. The Agent will check with the trader workstation 25 to make sure that the quote is still available. The Agent will send a HitAmountWS message to the workstation. The workstation will reply with a HitAmountWK message to show that the workstation is still working and that the trader did not interrupt the quote. At this point, the trader can 30 no longer interrupt the deal.

The Trading Agent will next check for available credit with the counterparty. The credit check may allow the deal, reduce the amount of the deal or disallow the deal. The Agent will then reduce the available credit by the amount needed for the deal. This reduction in available credit may affect future deals. The maker's Trading Agent will now inform the taker's Trading Agent of the deal by sending a 5 DealStatU3Maker message to its Broker node. The message is targeted to the identity of the hit. The network Broker nodes will route the message to the owner Broker node of the hit, and that Broker node will deliver it to the taker's Aaent. Once this message is sent, the maker's Agent knows 10 that a deal may have been done, but the deal is in doubt pending a reply. The taker's Trading Agent completes the deal execution process. This part of the process takes place when the Agent receives the DealStatusMaker message from the maker. If the message shows a valid deal, the process is continues.

The taker's Trading Agent will next check for available credit with the counterparty in a similar manner as the maker. The credit check may allow the deal, reduce the amount of the deal or disallow the deal. The Agent will then 20 reduce the available credit by the amount needed for the deal. This reduction in available credit may affect future deals. It should be remembered that deals are unlikely to be rejected at this stage as prices shown to traders are pre-screened for credit. The taker's Trading Agent will now 25 log the deal to its disk. As soon as the information is committed to persistent storage, the deal is done. Any checks on the deal status will now show a binding deal. The agent will now notify the trader, print a deal ticket and perform any other post deal processing. At this point, the 30 deal is done but the maker doesn't yet know. As soon as the deal is done, the taker's Trading Agent will notify the maker by sending a DealStatusTaker message to its Broker node. This message is targeted to the quote and will be routed to the maker's Agent.

The DealStatusTaker message contains final information about the deal, and therefore the final changes to the quote. This information is used by the network Broker nodes and the Trading Agent. As the DealStatusTaker message is 5 routed through the Broker nodes, each routing Broker node will use the information to update its quote context. The amount of the deal is moved from "reserved" to "complete".

The portion not done is moved from "reserved" to "available" if the quote is still active. It will then notify other 10 Broker nodes of the changes and of the deal by sending a MarketUpdate message to all other Broker nodes Using network routing rules.

When the DealStatusTaker message gets to the owner Broker node of the quote, it will send it to the Trading 15 Agent. The Agent will record the deal to disk. At this point the deal is no longer in doubt. The Agent will notify the trader, print a ticket and perform any other processing that is required. Some trading instruments may require additional information to be exchanged for a deal. An example of this 20 is the settlement instructions for EBS spot F/X. This type of information is sent in a DealInformation message. After the deal is processed, the Agents can develop this information. The DealInformation message is sent to the Broker node. The network Broker nodes will then route the 25 message to the other Agent where the information is processed as required by the instrument. A deal is thus completed.

once the deal is complete, the two parties will know the identity of their respective counterparty for the first 30 time. The identity will be displayed on theit terminal screen and shown, for example, in a listing of deals performed in that trading session as well as printed on the deal ticket and logged to disk. Each of these comprises a means for identifying to each of the parties to an executed 35 deal the counterparty to the deal.

The manner in which credit is handled in the system described will now be considered in more detail.

As mentioned previously, the system screens prices and matched deals using credit, as a result of which all prices 5 shown to a dealer should be available for trading. It will be understood from the foregoing description that this requires each broker to have sufficient credit information to be able to make credit decisions. This is because the broker nodes are responsible for forming the market view

10 which is distributed to communicating trading agents. The actual credit data is very complex and can vary by product and institution. For example, the concept of credit in an FIX trading system is straightforward as it is a spot market. However, for a product such as FRAs it is more is complex as deals are done over a variety of time periods. In addition, some banks may prefer to assign credit to a counterparty over the whole of the range of their trading activities whereas some banks will prefer to assign credit to counterparties for a given financial instrument.

To simplify the process the system distributes and uses a simple subset of the credit data. Final credit authority remains with a node that has the full credit information. in the present System this will be the banks trading agent 25 node but this is not mandatory and the invention is equally applicable to systems where credit is stored off site.

The system uses a single numeric value for each combination of trading floor, counterparty trading floor and tradable element. The purpose of the numerical value is to 30 determine whether the two floors have credit to deal in a particular element. The meaning of the numerical value is specific to the instrument being traded. For example, spot FIX uses the value as a yes/no flag (1 or 0) whereas in Forward Rate Agreements (FRA) the value is used as a bit mask for FRABBDA/ISDA decisions. Other instruments will have other meanings. The credit is bi-lateral. Credit must exist between two floors for any dealing activity to take place. The credit check is made for a given trading element 5 or pattern of trading elements as determined by the instrument. As the system is bilateral the broker will. compare two credit values; that given by the first floor to the second and that given by the second floor to the first. If the values are compatible, the dealing operation is 10 allowed. The meaning of compatible will be determined by the instrument. In terms of spot F/X if the amount proposed for the trade is lower or equal to the lowest of the two credit values the deal can proceed. Even if the deal is greater than the lowest credit value it may still proceed is but only for a part of the proposed deal amount equal to the lowest credit value.

The full credit information for a trading floor is originated for a trading agent that has credit authority for a trading floor. This agent only has part of the total 20 information; that relating to its own trading floor although it is possible that more that one trading floor is connected to a Trading Agent. When the credit information changes, the Trading Agent will sent a CreditUpdate message to its broker node. The broker will combine the information from 25 the Agent into its total credit matrix and pass the message to neighbour brokers as a broadcast message following the rules set out earlier. Each broker will also store a record of from where the credit information for a given floor came from.

30 In the prior art system described in W093/15467 the bank node holds the credit authority for a floor and is also responsible for dealing activity for the floor. The deal execution process described earlier is based on this credit model which is known as local credit.

During the deal execution the Trading Agent is presented with a potential deal. The Agent will examine the details of the deal and determine how much credit is required to complete the deal. It will check the available 5 credit and, if it is insufficient, the Agent may reduce the amount of the deal or disallow the deal. The amount of credit actually needed (the whole or reduced amount) is reserved from the pool of available credit. This credit is not available for other deals. If this reduces the 10 available credit for other deals below the dealing threshold the Agent will send a CreditUpdate message to notify the broker that credit is no longer available. If the reserved credit in due course becomes unreserved, the system may send a further CreditUpdate message to indicate that credit has is become available.

When the deal is completed, the maker's Agent will be notified with a DealStatusTaker message. The Taker's Agent will then be aware of the completed deal. The Agent will then determine the credit that was actually used by the 20 deal. This credit will be removed from the credit pool as consumed credit. Any remaining amount from the original reservation will be returned to the original pool.

As an alternative to local credit, a bank may adopt a Global Credit Model in which the Trading Agent that holds 25 the credit authority for a floor is not the same Agent that performs the dealing activity for that floor. The Agent with credit authority may, but does not have to, perform dealing activity for a floor. This arrangement allows all the floors of an institution to share a common pool of 30 credit and the creation of separate credit nodes within the network for some floors. The decision whether to adopt a global or local credit relationship is taken by participating banks and is external to the trading System.

It is essential that the system can deal with any 35 combination of local and global credit handling between counterparties. The various possibilities will now be Site 1 site 1 Site 2 site 2 Site 3 Site 3 Site 4 Site 4 described. In the following examples, Institution A is extending credit to Institution B. Example 1. Local - Global 5 Here Institution A operates under a local credit regime and Institution B under a global credit regime. Thus, each trading floors or Institution A must set a credit limit for deals with all floors in Institution B. 10 Institution A Institution B EXAMPLE 1: LOCAL - GLOBAL Local - Global relationships are common in the existing anonymous trading systems described previously, partly 15 because of the limitations of those Systems which require institutions to allocate credit limits between individual trading floors and each counterparty with which they want to trade. This has led to some banks allocating spot FX credit lines based on this grouping relationship to bring them into 20 line with the inflexible requirements of the prior art Systems. Other banks have set aside portions of their credit and allocated it using local - global credit purely for trading on anonymous systems - Such banks Use other credit relationships for their remaining operations.

Institution A Institution B Site 1 Site 1 Site 2 Site 2 Site 3 Site 3 Site 4 Site 4 Example 2: Global Global In this model, Institution A extends credit for all its 5 trading floors to all trading floors of Institution B. This means that the amount of trading any given floor of Institution A can do with Institution B is affected by the amount of trading with Institution B the other trading floors of Institution A have performed.

10 EXAMPLE 2: GLOBAL - GLOBA.L Global-Global credit relationships are the most fundamental or basic type.

Example 3: Global - Local In this model, the whole of Institution A extends a combined 15 line of credit to individual floors of Institution B. this may be appropriate where Institution A considers that Institution B is not uniformly trustworthy; there may be one particular trading floor with whom they are reluctant to - Institution A Institution B Site 1 Site 1 Site 2 Site 2 Site 3 Site 3 Site 4 Site 4 extend large lines of credit. Alternatively, the country in which a trading floor is located may need to be considered as a special case or a trading floor may not have an appropriate legal status within Institution B. 5 Example 3: GLOBAL-LOCAL Example 4: Local-Local In some situations, Institution A may allow individual trading floors to extend lines of credit to individual trading floors of Institution B. this occurs for example, in 10 emerging markets or exotic locations where trading is done solely from the branch of Institution A in that country to the branch of Institution B in that country. Typically, there is no other interaction between Institution A and these branches of Institution B. 15 Institution A Institution B Site 2 Site 2 Site 3 Site 3 Site 4 Site 4 EXAMPLE 4: LOCA-L-LOCA.L 5 When setting credit limits with a counterparty, an institution Must first define credit relationships and then allocate credit according to those relationships. These groupings may be done centrally. A typical credit grouping is as follows:

Ins titut ion B SITE 1 SITE 2 SITE 3 SITE 4 Institution B [Institution A Site 1 10 10 10 Site 2 10 10 10 Site 3, Site 4 12 is 12 is Table 1

In table 1 above, Institution A has created three groups of its own sites and three groups of Institution B sites. The Institution A groups are:

Site 1 --- --40- Site 1 20 Group 1 Site 1 Group 2 Site 2 Group 3 Sites 3 & 4 The Institution B groups are:

Group 1 site 1 Group 2 sites 2 & 3 Group 3 Site 4 The credit limits allocated are based on the group relationships. Thus, Group A3 extends $15M credit to Group B2 and $12M credit to Group B3. Table 1 may be written as follows:

B Group 1 Group 2 Group 3 A 10 Group 1 10 10 10 Group 2 10 10 10 Group 3 12 15 12 Table 2

Thus, Groups A3 and B2 are made up of more than one trading 15 floor and share single pools of credit. It will be appreciated that the group relationship shown is a hybrid of the local and global credit models. In the relationship between Institution A and Institution B, there may exist a plurality of local-global type relationships. Thus, Group A3 20 and Group B2 may represent all trading floors in major trading countries. The relationship between these groups is essentially Global- Global. On the other hand, the relationship between groups Al and Bl or B4 is a local-local relationship, the relationship between groups Al or A2 and 25 B2 is local-global and the relationship between-groups A3 and Bl or B3 is global-local. Other relationships are also present.

The above discussion refers to the relationship between Institutions A and B. A variety of relationships may also exist within the relationship between two Institutions. The flow of activity related to the creation of groups will depend on the following factors:

1. Grouping of the Institution sites; this may be 5 done:

a. Centrally - The Head Office defines all the groups for its sites or allows a proxy site to define the groups.

b. Distributed - The Head Office defines 10 certain main sites responsible for grouping themselves and other sites. For example, European sites may be grouped from London and US sites from New York. The authority for grouping sites may be based on 15 geography and/or legal status. In a distributed system certain groupings may still be handled centrally. For example if given local centre may have permission to group all sites in a given geographical 20 area with the exception of certain legal statuses which are handled centrally.

2. Authority is given for defining credit limits.

This may be:

a. Central 25 b. Distributed.

As in the grouping of the institution's own sites.

3. Counterparties must be grouped. Again, this may be done in the same manner as the in'stitutions 30 own sites are grouped. These three deCi3ions need not be implemented in the same manner.

A fully Global-Global relationship can exist between two institutions. There ail decision making is made centrally. In this case the credit grouping of tables 1 and 2 would be as follows:

Site 1 Site 2 Site 3 Site 4 Site I HO Site 2 99 Site 3 Site 4 Table 3

It should be understood that the above discussion may 10 relate to the manner in which banks allocate credit for a given financial instrument to be traded, for example, the allocation of credit for F/X spot trading. Depending on the nature of a given institution's own internal credit system, the credit relationship described may have to interface with 15 the institution's credit limits for the instrument being traded over all Possible methods including other anonymous systems, voice brokers and conversational dealing system.

Additionally or alternatively, it may be required to interface with an institution's global credit mechanism 20 which assigns credit to counterparty institutions across the range of its trading activities. Such a system is not confined to a single instrument but covers a range of instruments. Such systems are necessarily very complex as the notion of credit may vary where some instruments being 25 traded require credit to be extended over long periods of time.

Referring back now to the handling of global credit within the system of figures 1 to 7, it will be appreciated that the deal execution process for the global type of 30 credit arrangement is more complicated than for the local credit example described earlier.

The relationship between institutions many not be pure and hybrids may exist. For example, table 4 below shows a mixed Global-Global and Global local relationship between two institutions.

A B Site 1 Site 2 Site 3 Site 4 Subsidiary Site I HO 5 Site 2 so 10 Site 3 Site 4 Table 4

In this arrangement the head office decides its 10 own groupings, the counterparty groupings and the credit line. However, one subsidiary of Institution B is accorded credit separately from the remainder of the institution.

In table five below, Institution A maintains a locallocal relationship with a counterparty's floor, and still 15 allocates credit for most of the counterparty's floors on a global basis. Thus, Site 1, the head office of Institution A decides its own grouping, the counterparty grouping and the credit lines form Sites 1 to 4 but Site 5, which may be an exotic location has a local-local relationship with 20 Institution B site 4. Institution A Site 5 decides the credit line to Institution B, Site 4.

A B Site 1 Site 2 Site 3 Site 4, Exotic E Loc x Branch Site 1,HO Site 2 0 Site 3 Site 4 so Site 5, Exotic Loc x Branch 30 Table 5

A local-global relationship may have different credit lines for each of its f loors to the s ame global counterparty. This is not strictly a combination and is shown in Table 6 below. Site 1, the Head Office decides its own groupings, the counterparty groupings, but each Site administers the credit lines. (The lines themselves may be decided at the Head office).

site I Site 2 Site 3 Site 4 Site 1, HO is 10 Site 2 10 Site 3 10 Site 4 10 Site 5,small branch 5 Table 6 is Figure 8 shows the credit messageflow during deal execution with global credit.

The credit distribution process is the same as in the local credit example in that credit information is still distributed to all brokers. Each broker knows where the 20 information came from and can route a message back to the Trading Agent with credit authority.

In the example of figure 8, the Maker and Taker Trading Agents 100, 102 do not have credit authority for their floors. Credit must therefore be confirmed by the two 25 Trading Agents 120, 130 which do have that authority and which may be referred to as Maker and Taker Credit Agents.

When the Maker Trading Agent 100 processes a deal it will first check that the quote is still available in the manner described previously and it notifies the dealer of the pending deal. However, it cannot check the credit 5 position itself and so does not send the DealStatusMaker message itself. Instead, a DealCreditMaker message 140 is sent to the broker 150 to which the Trading Agent is attached. The broker 150 routes the DealCreditMaker message 140 to the Maker Credit Agent 120, which is the source of 10 credit information for the trading floor to which the Trading Agent 100 is performing the dealing activity. once the Maker Credit Agent 120 has performed the credit check as described previously, it sends the DealStatusMaker message to broker 170.

is The DealStatusMaker message 160 is routed by the broker not to the Taker Trading Agent but to the source of credit for the taker, in this case the two are not the same and the DealStatusMaker message is routed to the Taker Credit Agent 130. The Taker Credit Agent 130 then performs 20 credit checking as described previously and sends a DealCreditTaker message 180 to the broker 190 to which the Taker Credit Agent is connected. If the Taker Trading Agent has credit information for the trading floor the DealCreditTaker message 180 is not necessary.

25 The DealCreditTaker message 180 is routed by the broker network to the source of the original hit using the targeted routing rules described previously.

When the Trading Agent 110 that originally proposed the deal receives the DealCreditTaker message 180 the deal is 30 done and logged at the Taker Trading Agent and the deal execution process carries on as described earlier with respect to figure 6.

The Maker and Taker Credit Agents 120, 130perform credit reservation in the same manner as described in the local credit example. The Maker Credit Agent reserves credit when it receives the DealCreditMaker message and the 5 Taker Credit Agent reserves the credit when it receives the DealStatusMaker message 160. Credit consumption is then performed when the Maker and Taker Credit Agents 120, 130 receive the DealStatusTaker message 180 from the Taker Trading Agent 102.

10 It may be desired for more than one Trading Agent to hold the credit authority for a floor to increase reliability and performance. In such a case, any one such Credit Agent may confirm a deal. It is the responsibility of those Agents to communicate and keep the credit pool correct between themselves. This process is specific to an instrument or institution. Each broker will receive multiple CreditUpdate messages for the same floor. The brokers must decide which message to accept. The broker will examine a "hop count" in the message to determine which 20 message came from the closest source. The message with the higher hop count is not processed and is not routed.

The Credit Agent for a floor or institution has to maintain the pool of available credit and adjust the credit information as credit is used and restored. The manner in 25 which this is done is specific both to the institution and the instrument being traded.

One reason for a bank adopting a global approach to credit is to increase the flexibility available in trading. If a bank comprises several floors each of which have a 30 preassigned amount of credit with various counterparties, a situation can arise in which some of the floor trade up to their credit limits but others do not. Those floors which went up to their limits would have liked access to the unused credit on the other floors to maximise trading within the bank' s overall trading limit with a given party. That overall trading limit may not be confined to a single trading instrument but cover the range of the bank's activities, some of which may be traded on anonymous 5 electronic systems and others of which may not.

It will be appreciated that the system described allows any combination of local-global type credit relationships to be used. This enables institutions to assign credit to io operations using anonymous trading systems in the same manner as they do other trading operations. This enables a more flexible approach to credit to be adopted, where appropriate, compared to prior art systems. This in turn allows institutions more chance of trading up to their is credit limits and so maximising their profits from trading activities.

The credit limit relationship has been described with specific reference to a system using a network of broker nodes. It is to be understood that the invention is equally 20 applicable to any system which requires the allocation of bilateral credit limits between parties, including the systems described in EP-A-0,399,850, EP-A-0,406,026, EP-A-0,411,748 and W093/15467 referred to previously.

Various other modifications and alternative 25 arrangements to those described are possible and will occur to those skilled in the art. Such variations and modifications are all within the spirit and scope of the invention which is defined solely by the claims appended hereto.

Claims

What is claimed is:

1. An anonymous trading system for trading instruments between traders; comprising:

a communications network for transmitting s electronic messages; a plurality of order input devices connected to the communications network each for generating electronic order messages including bid and/or offer orders and for communicating to a trader order 10 information received from others of said plurality of order input devices over the network; at least one matching engine connected to the network for matching bid and offer orders input into the system from the trader terminals and for assisting is in executing deals where orders are matched; market distribution means connected to the network for distributing order messages to the trader terminals, the market distribution means being responsive to the order messages and the matching 20 engine; and credit limit storage means for storing credit limits available for trades between a trading floor or group of trading floors and possible counterparty trading floors or groups of trading floors and 25 comprising at least one credit agent node for storing credit limits for a group of logically separate trading floors.

2. An anonymous trading system according to' claim 1, wherein the order input devices for a given trading 30 floor are connected to a trading agent node connected to the communications network, wherein the credit agent node stores credit limits for trading floors connected to a plurality of trading agents.

3. An anonymous trading system according to claim 2 wherein the trading agent nodes having trading floors 5 connected whose credit limits are stored at the credit agent node each comprise means for sending credit enquiry messages to the credit agent node on receipt of a message that a bid or other input from one of said connected order input devices has been matched, and 10 wherein the credit agent node comprises means for receiving credit enquiry messages from each of the trading agent nodes for whose connected trading floors it stores credit limits, means for checking the proposed deal amount against stored credit limits on 15 receipt of a credit enquiry message and means for sending a message indicating whether or not the deal can proceed.

4. An anonymous trading system according to claim 3, wherein the credit enquiry messages to the credit agent 20 node via at least one of the plurality of matching engines.

5. An anonymous trading system according to claim 3 or 4, wherein the credit agent nodes sends the deal proceed indicator message to one of said matching engines and 25 said plurality of matching engines route said indicator messages to the trading agent node to which the counterparty trading floor is connected.

6. An anonymous trading system according to claim 3, 4 or S, wherein the means for sending a deal proceed 30 indicator message includes means for indicating that a deal may proceed only for a portion of the proposed amount.

7. An anonymous trading system according to claim 1, comprising means for interfacing the credit limit storage means of a given party to an external credit limit storage means of that party.

5

8. An anonymous trading system according to claim1 comprising a plurality of said credit agent nodes, each storing credit limits for a group of trading floors.

9. An anonymous trading system according to claim 1, wherein the credit limit storage means for each trading 10 floor or group of trading floors stores credit limits for individual counterparty trading floors, groups of counterparty trading floors, or a combination of individual and groups of counterparty trading floors.

10. An anonymous trading system according to claim 1, 15 comprising a plurality of interconnected broker nodes, each broker node comprising one of said plurality of matching engines and said market distribution means.

11. An automated trading system according to claim 1, wherein the bid and offer orders are input into the 20 system as price quotation messages and said market distribution means distributes price messages in response to said price quotation messages and the matching engine.

12. An automated trading system according to claim 1, 25 wherein the order input devices comprise trader terminals.

13. An automated trading system for anonymous trading of instruments, comprising a computer communications network having a plurality of interconnected broking 30 nodes, and a plurality of order input devices grouped in trading floors of one or more terminals, each trading floor being connected to a broker node through a trading agent node and each broker node comprising a matching engine for matching bid and offer orders input into the System from order input devices, means for 5 executing deals where orders are matched, and market distribution means for distributing order price messages to the order input devices in response to input bid and offer orders and the matching engine, the system further comprising at least one credit agent 10 node for storing credit limits for trading floors connected to a plurality of trading agent nodes for trades with possible counterparty trading floors or groups of trading floors.

14. An automated trading system according to claim 13,

15 wherein the bid and offer orders are input into the system as price quotation messages and said market distribution means distributes price messages in response to said price quotation Messages and the matching engine.

20 15. An automated trading system according to claim 13, wherein the order input devices comprise trader terminals.

16. An automated trading system for anonymous trading of instruments, comprising a computer communications 25 network having a plurality of interconnected broking nodes, and a plurality of order input devices grouped in trading floors of one or more devices, each trading floor being connected to a broker node through a trading agent node and each broker node comprising a 30 matching engine for matching bid and offer orders input into the system from order input devices, means for executing deals where orders are matched, and market distribution means for distributing order price messages to the order input devices in response to input bid and offer orders and the matching engine, the system further comprising credit limit storage means for storing credit limits for trading floors connected to a plurality of trading agent nodes for trades with 5 possible counterparty trading floors or groups of trading floors.

17. An anonymous trading system according to claim 16, comprising a plurality of said credit limit storage means, each storing credit limit for trading floors io connected to a plurality of trading agent nodes for trades with possible counterparty trading floors or groups of trading floors.

18. An automated trading system according to claim 16, wherein the bid and offer orders are input into the 15 system as price quotation messages and said market distribution means distributes price messages in response to said price quotation messages and the matching engine.

19. An automated trading system according to claim 16, 20 wherein the order input devices comprise trader terminals.