US-12620027-B2 - Message transmission timing optimization

US12620027B2US 12620027 B2US12620027 B2US 12620027B2US-12620027-B2

Abstract

Systems and methods are provided for message transmission timing optimization. The system receives a large market disrupting transaction and provides a transaction placement strategy that splits up the transaction to maximize favorable execution value and minimizes time required to execute the entire transaction.

Inventors

Inderdeep Singh
Frederic Cedric Malabre
Ari L. Studnitzer
David John Geddes

Assignees

CHICAGO MERCANTILE EXCHANGE INC.

Dates

Publication Date: 20260505
Application Date: 20210415

Claims (13)

1 . A computer implemented method comprising: preprocessing, by an order execution processor coupled with a database and a hardware matching processor, historical transaction and state data stored in the database to generate normalized historical transaction and state data defining a plurality of quantitative features; training, by the order execution processor, using the normalized historical transaction and state data, a machine learnt network coupled therewith, based on identification of one or more patterns in the normalized historical transaction and state data indicative of expected changes in the historical transaction and state data as a result of the historical transaction data in an electronic data transaction processing system to generate a trained machine learnt network, wherein the trained machine learnt network is a structured neural network comprising a layered plurality of interconnected processing nodes, wherein each connection of the plurality of interconnected processing nodes to another is dynamically weighted, wherein the electronic data transaction processing system is operative to transact a plurality of data items via the hardware matching processor associated therewith that matches, as they are received, incoming electronic data transaction request messages, for one of the plurality of data items received over a data communication network with at least one other previously received but unsatisfied electronic data transaction request message counter thereto stored in a memory associated with the plurality of data items coupled with the hardware matching processor, to at least partially satisfy one or both of the incoming electronic data transaction request messages or the at least one other previously received electronic data transaction request message; determining, by the order execution processor, using the trained machine learnt network, a predefined size threshold that would cause a change in the electronic data transaction processing system resulting in a disruption to the electronic data transaction processing system; determining, by a disruption filter processor coupled with the order execution processor, using the trained machine learnt network, that an incoming electronic data transaction request message exceeds the predefined size threshold that would cause the change in the electronic data transaction processing system resulting in the disruption thereto; predicting, by the order execution processor using the trained machine learnt network coupled therewith, based on an initial state of the electronic data transaction processing system, an expected change in the initial state as a result of one or more potential electronic data transaction request messages with one or more different parameters including at least a value and a transmission timing; generating, by an order optimizer processor, based on the prediction from the trained machine learnt network, before the incoming electronic data transaction request message that exceeds the predefined size threshold is processed, a transaction execution strategy comprising one or more alternative electronic data transaction request messages based on the incoming electronic data transaction request message that mitigate the disruption while maximizing one of the one or more different parameters and minimizing another, different from the one parameter, of the one or more different parameters, each of the one or more alternative electronic data transaction request messages not exceeding the predefined size threshold, the one or more different parameters selected based on the predicted expected change to minimize queue saturation of the hardware matching processor and maintain deterministic order matching; transmitting, by the order execution processor, the one or more alternative electronic data transaction request messages to the electronic data transaction processing system for processing in lieu of transmitting the incoming data transaction request message that exceeds the predefined size threshold; and processing, by the hardware matching processor, each of the one or more alternative electronic data transaction request messages in lieu of processing the incoming data transaction request message that exceeds the predefined size threshold based on the transaction execution strategy to dynamically control message flow and thereby improve operational stability, determinism, and throughput of the electronic data transaction processing system.
2 . The computer implemented method of claim 1 , further comprising: predicting, by the order execution processor using the trained machine learnt network responsive to receipt of another incoming electronic data transaction request message and based on a subsequent state of the electronic transaction processing system, expected further changes in the subsequent state as a result of the one or more potential electronic data transaction request messages with the one or more different parameters; generating, by the order optimizer processor using the prediction of the further expected changes, another alternative electronic data transaction request message that mitigates the disruption while maximizing the one parameter of the one or more different parameters and minimizing the other parameter of the one or more different parameters; and transmitting, by the order execution processor, the other alternative electronic data transaction request message to the electronic data transaction processing system.
3 . The computer implemented method of claim 1 , wherein the determining comprises: comparing a quantity value of the incoming electronic data transaction request message with the predefined size threshold; and determining that the incoming electronic data transaction request message is disruptive based on the comparison.
4 . The computer implemented method of claim 1 , wherein the plurality of quantitative features comprises at least an inside market width, a volume at a top of a market, a liquidity in the market, and a time window of observation.
5 . The computer implemented method of claim 1 , wherein the generating of the one or more electronic data transaction request messages further comprises: identifying a plurality of potential electronic data transaction request messages to be transacted sequentially within a time window, wherein the plurality of potential electronic data request messages together comprises a total quantity equal to a quantity stored in the incoming electronic data transaction request message, and wherein the plurality of electronic data transactions are identified based on the predictions to be transacted to maximize the value.
6 . The computer implemented method of claim 5 , wherein the plurality of electronic data transactions are further identified based on the predictions to minimize the timing.
7 . A system comprising: a database configured to store historical transaction and state data of an electronic data transaction processing system operative to transact a plurality of data items via a hardware matching processor associated therewith that matches, as they are received, incoming electronic data transaction request messages, for one of the plurality of data items received over a data communication network with at least one other previously received but unsatisfied electronic data transaction request message counter thereto stored in a memory associated with the plurality of data items coupled with the hardware matching processor, to at least partially satisfy one or both of the incoming electronic data transaction request messages or the at least one other previously received electronic data transaction request message; an order execution processor coupled with the database, the order execution processor configured to: preprocess the historical transaction and state data stored in the database to generate normalized historical transaction and state data defining a plurality of quantitative features; train, using the normalized historical transaction and state data, a trained machine learnt network, by identifying one or more patterns in the normalized historical transaction and state data indicative of expected changes in the historical transaction and state data as a result of the historical transactions; and determine a first predefined size threshold that would exceed a second predefined size threshold that would cause a change in the electronic data system causing a disruption to the electronic data transaction processing system and predict a future state of the electronic data transaction processing system as a function of a current state of the electronic data transaction processing system and one or more potential electronic data transaction request messages of different parameters, wherein the trained machine learnt network is a structured neural network comprising a layered plurality of interconnected processing nodes, wherein each connection of the plurality of interconnected processing nodes to another is dynamically weighted; a disruption filter processor configured to, based on predictions from the trained machine learnt network, identify that an incoming electronic data transaction request message exceeds the first predefined size threshold; and an order optimizer processor coupled with the trained machine learnt network and configured to, before the incoming electronic data transaction request message that exceeds the first predefined size threshold is processed, generate a transaction execution strategy comprising one or more alternative electronic data transaction request messages based on predictions by the trained machine learnt network, each of the one or more alternative electronic data transaction request messages not exceeding the first predefined size threshold, each of the one or more alternative electronic data transaction request messages maximizing one parameter and minimizing another parameter, different from the one parameter and not exceeding the first predefined size threshold, the one or more different parameters selected based on the predicted expected change to minimize queue saturation of the hardware matching processor and maintain deterministic order matching, wherein the order execution processor is further configured to transmit the one or more alternative electronic data transaction request messages to the electronic data transaction processing system for execution in lieu of transmission of the incoming data transaction request message, wherein the hardware matching processor of the electronic data transaction processing system is configured to execute the alternative electronic data transaction request messages in lieu of execution of the incoming data transaction request message that exceeds the first predefined size threshold based on the transaction execution strategy to dynamically control message flow and thereby improve operational stability, determinism, and throughput of the electronic data transaction processing system.
8 . The system of claim 7 , wherein the plurality of quantitative features comprises at least an inside market width, a volume at a top of a market, liquidity in the market, and a time window of observation.
9 . A computer implemented method comprising: preprocessing, by an order execution processor coupled with a database, historical transaction and state data stored in the database to generate normalized historical transaction and state data defining a plurality of quantitative features; training, by the order execution processor, a machine learnt network coupled therewith, based on identification of one or more patterns in the normalized historical transaction and state data indicative of expected changes in the historical transaction and state data as a result of the historical transaction data in an electronic data transaction processing system to generate a trained machine learnt network, wherein the trained machine learnt network is a structured neural network comprising a layered plurality of interconnected processing nodes, wherein each connection of the plurality of interconnected processing nodes to another is dynamically weighted; determining, by the order execution processor, using the trained machine learnt network, a predefined size threshold that would cause a change in the electronic data transaction processing system resulting in a disruption to the electronic data transaction processing system; identifying, by a disruption filter processor, an incoming transaction which exceeds the predefined size threshold that would cause the change in the electronic data transaction processing system resulting in the disruption thereto; predicting, by the order execution processor, using the trained machine learnt network, based on a calculated state of a data structure, an expected change in the state in the data structure for one or more potential transactions to be transacted at one or more values at one or more different times; generating, by an order optimizer processor, before the incoming transaction that exceeds the predefined size threshold is executed, based on the predictions from the trained machine learnt network, a transaction execution strategy comprising one or more alternative transactions, each of the one or more alternative transactions not exceeding the predefined size threshold, wherein the one or more alternative transactions maximize an execution value and limits a time to transact, the execution value and the time to transact selected based on the predicted expected change to minimize queue saturation and maintain deterministic order matching; transmitting, by the order execution processor, the one or more alternative transactions to a hardware matching processor of the electronic trading system for execution in lieu of transmission of the incoming transaction; and executing, by the hardware matching processor, the one or more alternative transactions in lieu of executing the incoming transaction that exceeds the predefined size threshold based on the transaction execution strategy to dynamically control message flow and thereby improve operational stability, determinism, and throughput of the electronic data transaction processing system.
10 . The computer implemented method of claim 9 , wherein the plurality of quantitative features comprises at least an inside market width, a volume at a top of a market, liquidity in the market, and a time window of observation.
11 . The computer implemented method of claim 9 , wherein the executing of the one or more alternative transactions further comprises transmitting the one or more alternative transactions to an electronic data transaction processing system in lieu of the incoming transaction.
12 . A system comprising: means for preprocessing historical transaction and state data stored in a database to generate normalized historical transaction and state data defining a plurality of quantitative features; means for training a machine learnt network coupled therewith, based on identification of one or more patterns in the normalized historical transaction and state data indicative of expected changes in the historical transaction and state data as a result of the historical transaction and state data in an electronic data transaction processing system to generate a trained machine learnt network, wherein the trained machine learnt network is a structured neural network comprising a layered plurality of dynamically weighted interconnected processing nodes, wherein each connection of the plurality of interconnected processing nodes to another is dynamically weighted, wherein the electronic data transaction processing system is operative to transact a plurality of data items via a hardware matching processor associated therewith that matches, as they are received, incoming electronic data transaction request messages, for one of the plurality of data items received over a data communication network with at least one other previously received but unsatisfied electronic data transaction request message counter thereto stored in a memory associated with the plurality of data items coupled with the hardware matching processor, to at least partially satisfy one or both of the incoming electronic data transaction request messages or the at least one other previously received electronic data transaction request message; means for determining, by the processor, using the trained machine learnt network, a predefined size threshold that would cause a change in the electronic data transaction processing system resulting in a disruption to the electronic data transaction processing system; means for determining that an incoming electronic data transaction request message exceeds the predefined size threshold that would cause a change in the electronic data transaction processing system resulting in a disruption thereto; means for predicting, using the trained machine learnt network, based on an initial state of the electronic data transaction processing system, an expected change in the initial state as a result of one or more potential electronic data transaction request messages with one or more different parameters including at least a value and a transmission timing; means for generating based on the prediction from the trained machine learnt network, before the incoming electronic data transaction request message that exceeds the predefined size threshold is processed, a transaction execution strategy comprising one or more alternative electronic data transaction request messages based on the incoming electronic data transaction request message that mitigate the disruption while maximizing one of the one or more different parameters and minimizing another, different from the one parameter, of the one or more different parameters, each of the one or more alternative electronic data transaction request messages not exceeding the predefined size threshold, the one or more different parameters selected based on the predicted expected change to minimize queue saturation and maintain deterministic order matching; means for transmitting the one or more alternative electronic data transaction request messages to the electronic data transaction processing system for processing in lieu of transmitting the incoming data transaction request message that exceeds the predefined size threshold; and means for processing, each of the one or more alternative electronic data transaction request messages in lieu of processing the incoming data transaction request message that exceeds the predefined size threshold based on the transaction execution strategy to dynamically control message flow and thereby improve operational stability, determinism, and throughput of the electronic data transaction processing system.
13 . The system of claim 12 , wherein the expected change in the state is predicted based on historical data comprising at least an inside market width, a volume at a top of a market, liquidity in the market, and a time window of observation.

Description

RELATED APPLICATIONS This application is a continuation under 37 C.F.R. § 1.53(b) of U.S. patent application Ser. No. 15/889,751 filed Feb. 6, 2018, now U.S. Pat. No. 11,023,969, the entirety of which is incorporated by reference herein and relied upon. BACKGROUND A financial instrument trading system, such as a futures exchange, referred to herein also as an “Exchange”, such as the Chicago Mercantile Exchange Inc. (CME), provides a contract market where financial products/instruments, for example futures and options on futures, are traded. Futures is a term used to designate all contracts for the purchase or sale of financial instruments or physical commodities for future delivery or cash settlement on a commodity futures exchange. A futures contract is a legally binding agreement to buy or sell a commodity at a specified price at a predetermined future time, referred to as the expiration date or expiration month. An option is the right, but not the obligation, to sell or buy the underlying instrument (in this case, a futures contract) at a specified price within a specified time. The commodity to be delivered in fulfillment of the contract, or alternatively, the commodity, or other instrument/asset, for which the cash market price shall determine the final settlement price of the futures contract, is known as the contract's underlying reference or “underlier.” The terms and conditions of each futures contract are standardized as to the specification of the contract's underlying reference commodity, the quality of such commodity, quantity, delivery date, and means of contract settlement. Cash Settlement is a method of settling a futures contract whereby the parties effect final settlement when the contract expires by paying/receiving the loss/gain related to the contract in cash, rather than by effecting physical sale and purchase of the underlying reference commodity at a price determined by the futures contract price. Some products on an exchange are traded in an open outcry environment where the exchange provides a location for buyers and sellers to meet and negotiate a price for a quantity of a product. Other products are traded on an electronic trading platform (e.g., an electronic exchange), also referred to herein as a trading platform, electronic trading system, trading host or Exchange Computer System, where market participants, e.g. traders, use software to send orders to the trading platform. The order identifies the product, the quantity of the product the trader wishes to trade, a price at which the trader wishes to trade the product, and a direction of the order (i.e., whether the order is a bid, i.e. an offer to buy, or an ask, i.e. an offer to sell). It will be appreciated that there may be other order types or messages that traders can send including requests to modify or cancel a previously submitted order. The speed in which trades are executed through electronic trading systems provide many benefits. Electronic trading systems can facilitate a large number of market transactions. The greater the number of market transactions, the greater a market's liquidity. In liquid markets, prices are driven by competition; prices reflect a consensus of an investment's value; and trading systems provide a free and open dissemination of information. With the advent of improved computational and communications capabilities, the speed and efficiency with which traders may receive information and trade in electronic trading systems has greatly improved. Algorithmic and high frequency trading utilize computers to quickly analyze market information and place trades allowing traders to take advantage of even the smallest movements in prices. Unfortunately, this improved speed and efficiency also improves the speed at which problems may occur and propagate, such as where the market ceases to operate as intended, i.e. the market no longer reflects a true consensus of the value of traded products among the market participants. Such problems are typically evidenced by extreme market activity such as large changes in price, whether up or down, over a short period of time or an extreme volume of trades taking place. In particular, traders, whether human or electronic, may not always react in a rational manner, such as when presented with imperfect information, when acting in a fraudulent or otherwise unethical manner, and/or due to faulty training or design. For example, while communications technologies may have improved, inequities in access to information and opportunities to participate still exist, which may or may not be in compliance with legislative, regulatory and/or ethical rules, e.g. some traders receive information before other traders, some traders may be able to place trader orders more quickly than others. In many cases, irrational trader behavior may be triggered by a market event, such as a change in price, creating a feedback look where the initial irrational reaction may then cause further ma