This document discusses the use of artificial intelligence in financial investing and identifies some potential pitfalls. It notes that while AI can help analyze large datasets and identify new opportunities, financial time series data can be non-stationary, and AI models may overfit data and lack proper out-of-sample testing. Subject matter expertise is still key to integrating new data sources like news articles. The document provides examples of AI applications at J.P. Morgan Asset Management in areas like trading, earnings estimates, and thematic portfolios.
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
AI's Impact on Financial Investing
1. FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
The Pitfalls of Using AI in Financial Investing
November 2018
Yazann Romahi, PhD
2. 1 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
Areas of AI Applications in Finance
Trading
Fraud Detection
Credit LendingImage Recognition
Robo-Advisors
For illustrative purposes only
$
Asset Management
3. 2 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
The March of Quantitative Methods in Financial Investing
Markowitz
Modern
Portfolio
Theory
Louis
Bachelier’s
Thesis
published
Rapid
Advances
in Portfolio
Theory
1980s
First CTA
funds
1990s
Growth of
Quant Equity
Long Short
Hedge Funds
2000s
Increasing Use
of AI in Trading
Strategies
2009-
Growth of
Alternative
Beta
Growth of Quantitative Investing Strategies
New sources
Of Data
Rapidly Being
Created
1900 1952 1960s 1980 2010 2014
𝒌=𝟎
𝒏
𝒏
𝒌
𝒙 𝒌
For illustrative purposes only
4. 3 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
Understanding Approaches to Building Investment Strategies
$370 Bn
$973 Bn
Hedge Fund and CTA AUM
Trend Following
Trend-following hedge funds (also known as
CTAs) are exclusively quantitatively based
Typically employ momentum and reversion
signals at different time frequencies
AI is used, but traditional methods are more
prevalent
Fundamental
Equity
Quantitative, equity long/short approaches
began to gain prominence in the early 90s.
Fundamental signals (e.g. valuation, quality) are
an important component of these processes
New sources of data are creating new
opportunities for alpha generation
Source: BarclayHedge as of 2Q 2018. For illustrative purposes only
5. 4 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
An Expansion of the Quant’s Toolkit
NEW TECHNIQUES
TRADITIONAL
DATA
NEWDATA
TRADITIONAL TECHNIQUES
Building a neural
network (machine
learning) on textual data
Building random
forests for fraud
detection
Time series extracted
from satellite data of
industrial sites in China
Regression on
Earnings Yield;
Time series analysis
using econometrics
For illustrative purposes only
6. 5 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
New Data Sources Abound
of the data available today has been
collected in the last two years90%
Alternative Data
Individuals Business Processes Sensors
Social Media Transaction Data Satellites
News and Reviews Corporate Data Geo-location
Web Searches,
Personal Data
Government Agencies Data Other Sensors
For illustrative purposes only
7. 6 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
Pitfalls of Time Series Analysis in Trading Strategies
New sources of data have cross-sectional depth, but lack time-series depth1
Probability of random noise ~30% Not statistically significant (p=0.12)
For illustrative purposes only
Monthly Trading Model (55% Success Rate) Weekly Trading Model (55% Success Rate)
8. 7 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
Pitfalls of Time Series Analysis in Trading Strategies
Financial time-series are non-stationary2
Stationary Data
-15%
-10%
-5%
0%
5%
10%
15%
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
S&P 500
(Daily Returns, 2008)
-1,000
-800
-600
-400
-200
0
200
400
600
1995
1997
1999
2001
2003
2005
2007
2009
2011
2013
2015
Total Nonfarm Payrolls
(Change, Thousands of Persons)
Source: Bloomberg, as of November 2018. For illustrative purposes only
9. 8 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09
S&P500Level
Example Financial Time Series
Pitfalls of Time Series Analysis in Trading Strategies
Everything should be made as simple as possible, but not simpler3
Traditional Methods
• Require modeler to determine functional form
• Econometrics methods are often perfectly adequate
Artificial Intelligence
• Significantly higher degrees of freedom allows
for flexibility, but is often less statistically robust
due to inability to properly test out of sample
Source: Bloomberg, as of November 2018. For illustrative purposes only
0
20
40
60
80
100
120
ModelError
Model Complexity
Model Complexity vs. Error
Total Model Error
In-sample Error
Best Model
10. 9 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
100
120
140
160
180
200
220
240
Jan-17 Feb-17 Mar-17 Apr-17 May-17 Jun-17 Jul-17 Aug-17 Sep-17 Oct-17
Subject Matter Expertise is Key: An Example
Corporate activity is a source of binary idiosyncratic risk
Stock price movements following confirmation
or denial of a merger agreement
News articles
are a source of
useful
information, if
appropriately
handled
For illustrative purposes only
11. 10 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
2% RELEVANT
NEWSMILLIONS OF
ARTICLES
PER YEAR1
Investment Process Application of AI: NewsFilter
1 For illustrative purposes only
2 Awarded based on use of machine-learning based News-Filter
Most Cutting-Edge IT Initiative
Best use of Emerging or Innovative Technology
AWARDS2
AI can provide a solution to managing large data sets more effectively allowing it to be systematically incorporated
in factor portfolios
12. 11 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
Investment Process Application of AI: NewsFilter
Machine learning algorithm can improve performance by reducing idiosyncratic risk
Source: J.P. Morgan Asset Management, Bloomberg
15
20
25
30
35
1-Sep 15-Sep 29-Sep 13-Oct 27-Oct 10-Nov 24-Nov 8-Dec 22-Dec
General Cable Corporation
~75% increase in
price post rumor
Identification of M&A rumor led to short constraint on General Cable in
equity factor models, preventing loss of 25bps at portfolio level
13. 12 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
Areas of AI Applications at J.P. Morgan Asset Management
Trading
Using reinforcement learning in
enhanced trading
Earnings Revisions
Analysis of company earnings reports
and Q&A yielding better estimates of
earnings
Image Recognition
Analysis of industrial site satellite
imagery in China as a leading
indicator of industrial production
Thematic Portfolios
Using NLP of news, social media and
financial reports, to create on demand
thematic portfolio ideas
$
14. 13 | FOR EDUCATIONAL PURPOSES ONLY | NOT INTENDED TO BE USED AS ADVERTISING OR SALES LITERATURE
Disclosures
The views contained herein are not to be taken as advice or a recommendation to buy or sell any investment in any jurisdiction, nor is it a commitment from J.P. Morgan Asset Management or any of
its subsidiaries to participate in any of the transactions mentioned herein. Any forecasts, figures, opinions or investment techniques and strategies set out are for information purposes only, based on
certain assumptions and current market conditions and are subject to change without prior notice. All information presented herein is considered to be accurate at the time of production. This material
does not contain sufficient information to support an investment decision and it should not be relied upon by you in evaluating the merits of investing in any securities or products. In addition, users
should make an independent assessment of the legal, regulatory, tax, credit and accounting implications and determine, together with their own professional advisers, if any investment mentioned
herein is believed to be suitable to their personal goals. Investors should ensure that they obtain all available relevant information before making any investment. It should be noted that investment
involves risks, the value of investments and the income from them may fluctuate in accordance with market conditions and taxation agreements and investors may not get back the full amount
invested. Both past performance and yields are not reliable indicators of current and future results.
J.P. Morgan Asset Management is the brand for the asset management business of JPMorgan Chase & Co. and its affiliates worldwide.
To the extent permitted by applicable law, we may record telephone calls and monitor electronic communications to comply with our legal and regulatory obligations and internal policies. Personal
data will be collected, stored and processed by J.P. Morgan Asset Management in accordance with our Company’s Privacy Policy. For further information regarding our regional privacy policies
please refer to the EMEA Privacy Policy; for locational Asia Pacific privacy policies, please click on the respective links: Hong Kong Privacy Policy, Australia Privacy Policy, Taiwan Privacy Policy,
Japan Privacy Policy and Singapore Privacy Policy.
This communication is issued by the following entities: in the United Kingdom by JPMorgan Asset Management (UK) Limited, which is authorized and regulated by the Financial Conduct Authority; in
other European jurisdictions by JPMorgan Asset Management (Europe) S.à r.l.; in Hong Kong by JF Asset Management Limited, or JPMorgan Funds (Asia) Limited, or JPMorgan Asset Management
Real Assets (Asia) Limited; in Singapore by JPMorgan Asset Management (Singapore) Limited (Co. Reg. No. 197601586K), or JPMorgan Asset Management Real Assets (Singapore) Pte Ltd (Co.
Reg. No. 201120355E); in Taiwan by JPMorgan Asset Management (Taiwan) Limited; in Japan by JPMorgan Asset Management (Japan) Limited which is a member of the Investment Trusts
Association, Japan, the Japan Investment Advisers Association, Type II Financial Instruments Firms Association and the Japan Securities Dealers Association and is regulated by the Financial
Services Agency (registration number “Kanto Local Finance Bureau (Financial Instruments Firm) No. 330”); in Australia to wholesale clients only as defined in section 761A and 761G of the
Corporations Act 2001 (Cth) by JPMorgan Asset Management (Australia) Limited (ABN 55143832080) (AFSL 376919); in Brazil by Banco J.P. Morgan S.A.; in Canada for institutional clients’ use
only by JPMorgan Asset Management (Canada) Inc., and in the United States by JPMorgan Distribution Services Inc. and J.P. Morgan Institutional Investments, Inc., both members of FINRA; and
J.P. Morgan Investment Management Inc. In APAC, distribution is for Hong Kong, Taiwan, Japan and Singapore. For all other countries in APAC, to intended recipients only.
Copyright 2018 JPMorgan Chase & Co. All rights reserved.
Notes de l'éditeur
Maybe put right after timeline?
Going to walk through main areas of quant use in HF land
Two main areas – find better icons? Ai – Smarter trend following, but traditional is adequate and upside for algos is limited
Fundamental – more domain knowledge necessary – equity long short 1/3 of AUM, fundamental signals at the core of the model – value – some version of cheapness. Analyst estimate earnings or trailing. But with new data, there are ways to look at earnings with new data sources – credit card data – Netflix, etc
Interesting applications
When AI researchers come to finance, need to rethink
AI are black box models
When you have functional forms, assumptions are cleaner and relationships are clear
Not enough data to adapt to non-stationary
Need to remove clicks
Use the above template as a starting point for your disclosure page(s). Product-specific risk disclosure may be obtained on fact sheets, or product web pages.
Regarding footnotes: It is NOT permissible to run footnotes together in large blocks of print. If disclosure is in a footnote, separate or number the footnotes so they can be read. Therefore, footnotes must be at least 8-pt font, and in the same font-type used in the majority of the presentation. Using smaller-type fonts, such as Calibri or Arial Narrow, is not permitted.