Machine Learning

in FX

August 8, 2019

J.P. Morgan is taking technology to a new level in the foreign exchange market, applying machine learning to provide competitive pricing and optimize execution in what is already one of the most liquid and automated asset classes alongside equities.

DNA is an optimization feature that leverages simulated data from various types of market conditions to select the best order placement and execution style designed to minimize market impact. Chi Nzelu Head of Macro eCommerce J.P. Morgan

The Deep Neural Network for Algo Execution (DNA) is J.P. Morgan's latest tool to enhance its FX algorithms and uses a machine learning framework to bundle certain existing algos into one streamlined execution strategy.

“DNA is an optimization feature that leverages simulated data from various types of market conditions to select the best order placement and execution style designed to minimize market impact," said Chi Nzelu, head of Macro eCommerce at J.P. Morgan. “It then uses reinforcement learning – a subset of machine learning – to assess the performance of individual order placement choices."

Next Generation Algos

Algorithms have undergone a series of evolutions since their birth in the equities and foreign exchange markets over a decade ago.

First-Generation

So-called first-generation algorithms consisted of buy or sell orders with simple parameters. The next wave of algorithms were more sophisticated, providing investors with dynamic pricing derived from mathematical theory.

Second-Generation

Second-generation algos deployed strategies to break up large orders and reduce potential market impact. For example, selling 500 million euros versus dollars in a short period of time could cause the price to sharply decline and therefore cost the investor more. However, slicing the order into small 'clips' could reduce market impact and help obtain a better price.

Time-Weighted Average Price

The time-weighted average price (TWAP) algo, as such, allows a user to select a time frame over which to execute the trade. A volume-weighted average price (VWAP) algo is another traditionally-used strategy where a time schedule is normalized by the expected distribution of volume.

Over the years, “strategies” such as these have multiplied and clients have been given the ability to choose from a suite of algos that offer various execution methods designed to achieve optimal execution quality with minimal market impact.

While DNA is currently an enhancement for certain existing strategies, the future goal is to create one, all-encompassing algorithm that uses available data to provide users with information to improve execution under various market conditions.

Behind Machine Learning

To create one algorithm with increased logical capacities, the strategists behind DNA used reinforcement learning. By using deep pools of data that simulate multiple market scenarios, reinforcement learning trains the algo to learn from the actions it takes.

This is a fundamental shift from early generation algos, which were primarily built off human-based programming or rule-based executions.

“Artificial neural networks (ANN), of which DNA is a type, are inspired by the biological neural networks of the brain. They are capable of modelling complex non-linear relationships with little restriction in the inputs, which is useful when trying to model reality because relationships in real life are often complicated,” said Sam Nian, a Lead Strategist in the DNA initiative.

Using the analogy of teaching a robot how to walk, rules-based technology would program the robot to lift one leg followed by another to move forward. More sophisticated technology – equivalent to that behind second-generation algorithms – would show the robot billions of videos that demonstrate how to walk. Reinforcement learning throws the robot into different environments and forces it to walk by learning the way a toddler would – by experience, by falling down, by running into obstacles.

Artificial neural networks (ANN), of which DNA is a type, are inspired by the biological neural networks of the brain. They are capable of modelling complex non-linear relationships with little restriction in the inputs, which is useful when trying to model reality because relationships in real life are often complicated. Sam Nian DNA Lead Strategist J.P. Morgan

“The robot developed by Alphabet Inc. that beat a human professional in the strategy game Go used reinforcement learning and a deep neural network similar to the one behind DNA. Instead of relying on statistical regression, supervised learning, and human hard-coded rules, the reinforcement learning approach provides more flexibility and removes potential human bias when training the model," said Tanya Tang, another Lead Strategist on the project.

Taking a Cue from Equities

While the foreign exchange market has been at the forefront of technology since the 1990s – when investment banks across the street developed platforms for clients to trade electronically rather than through voice traders – the idea to create DNA was inspired by technological developments in the equities space, according to the J.P. Morgan team.

J.P. Morgan rolled out a proprietary equities trading execution offering powered by machine learning in 2017 that optimizes between liquidity demand and passive trading, adapting as market conditions change.

Both offerings benefit from reinforcement learning techniques designed to optimize the resultant execution and consequently price by making a decision between a number of pre-defined market actions and strategies using historical and simulated data.

In order to create the most scenarios and simulated environments possible, J.P. Morgan developers selected G7 currencies because they are the most heavily traded and therefore have the most data to teach the machine. While still in the initial stage, DNA has demonstrated its ability to push the performance of J.P. Morgan algos to an even higher level.

Related Insights

FOR INSTITUTIONAL & PROFESSIONAL CLIENTS ONLY – NOT INTENDED FOR RETAIL CUSTOMER USE

This material has been prepared by J.P. Morgan Sales and Trading personnel and is not the product of J.P. Morgan’s Research Department. It is not a research report and is not intended as such.  This material is provided for informational purposes only and is subject to change without notice.  It is not intended as research, a recommendation, advice, offer or solicitation to buy or sell any financial product or service, or to be used in any way for evaluating the merits of participating in any transaction.  Please consult your own advisors regarding legal, tax, accounting or any other aspects including suitability implications for your particular circumstances.  J.P. Morgan disclaims any responsibility or liability whatsoever for the quality, accuracy or completeness of the information herein, and for any reliance on, or use of this material in any way.  This material is provided on a confidential basis and may not be reproduced, redistributed or transmitted, in whole or in part, without the prior written consent of J.P. Morgan.  Any unauthorized use is strictly prohibited.  The products and/or services mentioned herein may not be suitable for your particular circumstances and may not be available in all jurisdictions or to all clients.  Clients should contact their salespersons at, and execute transactions through, a J.P. Morgan entity appropriately licensed in the client’s home jurisdiction unless governing law permits otherwise.  This material is a “solicitation” of derivatives business only as that term is used within CFTC Rule 1.71 and 23.605. Where this material is an “investment recommendation” as that term is defined in MAR visit: www.jpmm.com/#mardisclosures. This material is subject to terms at: www.jpmorgan.com/salesandtradingdisclaimer.

 

© 2019 JPMorgan Chase & Co. All rights reserved. J.P. Morgan is a marketing name for investment banking businesses of JPMorgan Chase & Co. and its subsidiaries and affiliates worldwide.  Bank products and services, including certain lending, derivative and other commercial banking activities, are offered by JPMorgan Chase Bank N.A. (JPMCB), including through its authorized branches and other global affiliates registered with local authorities as appropriate.  Securities products and services, including execution services, are offered in the United States by J.P. Morgan Securities LLC (JPMS LLC), in EMEA by J.P. Morgan Securities plc (JPMS plc) where permitted and in other jurisdictions worldwide by other appropriately licensed global affiliates.  JPMCB, JPMS LLC and JPMS plc are principal subsidiaries of JPMorgan Chase & Co.  For information on which legal entities offer investment banking products and services in each jurisdiction, please consult: www.jpmorgan.com/ib-legal-entities.  For important disclosures in respect of securities transactions, please consult: www.jpmorgan.com/securities-transactions and in respect of over-the-counter equity derivatives transactions, please consult: www.jpmorgan.com/otc-equity-derivative-transactions

 

For additional regulatory disclosures, please consult: www.jpmorgan.com/disclosures.

 

Copyright © 2019 JPMorgan Chase & Co. All rights reserved.