Transcript Slide 1
Trading Off the News:
Applications of News Algorithms
July 1, 2008
CARISMA 2008
Alan Slomowitz
Director Product Development
Algorithmic and Trading Solutions
Dow Jones Content Technology Solutions
© Copyright 2008 Dow Jones and Company
|
Trading Off The News - HP to Acquire EDS
15:02 *WSJ: Hewlett-Packard In Advanced Talks to Acquire EDS
15:02 *WSJ: Hewlett-Packard-EDS Deal Could Come In Matter Of Days
15:03 *WSJ: Premium Expected To Be In 30% Range –Sources
15:03 *WSJ: EDS Overall Price Roughly $13 Billion -Sources >HPQ EDS
Source: DJ Newswires
© Copyright 2008 Dow Jones and Company
|
3
News and the Quantitative Investor:
The New Frontier
Successful trading is about finding the crucial edge
First to take advantage of trading opportunities
Quantitative investors require:
Analytical hypotheses
Technology
Data
Data = News
Exploit this new information source
Difficult-to-quantify qualitative content
© Copyright 2008 Dow Jones and Company
|
4
Taking Advantage of the News
Intuitively, investment professionals realize you
can use news to create profitable trading models
Proprietary traders, Stat Arb desks and hedge
funds now use news to create Alpha-based
trading strategies
News based algorithms that can profit from market anomalies lasting from mere
milliseconds to days or even longer
Economic indicator based algorithms
Corporate news based algorithms
News-based algorithms are valuable additions to
Alpha generation:
News is difficult to integrate into investment models
Difficulty means higher value – not so easy to commoditize
Requires different tools – both data and text analysis tools
Different ways of analyzing market trends
© Copyright 2008 Dow Jones and Company
|
Making News Work for You:
Tagged, Machine-Readable News Feed
Economic Indicator-Based Algorithms
Arms Race – Must have a global, reliable and innovative news
provider
Latency matters
FX and Futures markets
Multiple markets and countries
Corporate News-Based Algorithms
Scheduled events: Earnings, Guidance
Unscheduled events: Mergers, Ratings changes, Executive
changes
Text mining events: surprise and sentiment
Analyzing the language of the news
Quantifying that analysis
© Copyright 2008 Dow Jones and Company
|
Solving the Corporate News Puzzle
Expected Corporate News Events
Plan for releases incorporating expected, historical and actual data
How long does it take for market to assimilate this news
Unexpected News Events
Cannot build an alpha algorithm for every unexpected event
Anticipate a controlled set of pre-defined news events
Ratings changes, executive changes, dividends
Capability to stop algorithms that are in flight when such news events
occur
e.g., HP announces purchase of EDS
Adjust trading strategy for significant events
© Copyright 2008 Dow Jones and Company
|
Solving the Corporate News Puzzle (cont.)
Complex Event Processing
Multiple signals combined
If EPS > X and Price moves > 3% in 5 minutes and market is up less
than 1% Then BUY XXXX shares
Academic Research
© Copyright 2008 Dow Jones and Company
|
8
What the Academics are Saying
Paul Tetlock, Quantifying Language to Measure Firms’
Fundamentals (Sept 2006/May 2007)
Linguistic media content captures otherwise hard-to-quantify aspects
of firms’ fundamentals, which investors quickly factor into stock prices
Simple counting of negative words can uncover negative sentiment in
earnings releases
Antweiler and Frank, Do U.S. Stock Markets Typically
Overreact to Corporate News Stories? (October, 2005)
Initial returns jump is typically accompanied by a temporary jump in
trading volume
Reversal is typically accompanied by gradually declining trading
volume
During business cycle expansions, the process is largely complete
after two or three weeks
During recessions, the reversal is much more prolonged and we are
not able to identify a clear point at which the process ends
© Copyright 2008 Dow Jones and Company
|
9
What the Academics are Saying (cont.)
Paul Tetlock, Giving Content to Investor Sentiment: The
Role of Media in the Stock Market, Journal of Finance
(June 2007)
Academic Bibliography available
© Copyright 2008 Dow Jones and Company
|
Trading off Positive/Negative Word Counts
Long/Short Portfolio
• Firms with positive news –
Long
• Firms with negative news –
Short
• Hold for one trading day
• Rebalance at end of next
trading day
More Than Words – Quantifying Language to Measure Firms’ Fundamentals
May 2007 – Tetlock, Saar-Tsechansky, Macskassy
© Copyright 2008 Dow Jones and Company
|
Abnormal Event Returns 10 Days Prior and
10 Days Post News Releases
S&P 500 Firms
All news stories with a fraction of
negative words in the year’s top
(bottom) quartile as negative
(positive) stories
Compute difference between
reaction to positive and negative
news stories
More Than Words – Quantifying Language to Measure Firms’ Fundamentals
May 2007 – Tetlock, Saar-Tsechansky, Macskassy
© Copyright 2008 Dow Jones and Company
|
Cumulative Standardized Unexpected Earnings (SUE)
Cumulative SUE 10 quarters prior
to news and 10 quarters after news
Negative stories 30 days prior to
earnings announcements
More Than Words – Quantifying Language to Measure Firms’ Fundamentals
May 2007 – Tetlock, Saar-Tsechansky, Macskassy
© Copyright 2008 Dow Jones and Company
|
40 Minutes of Alpha
S&P 500 Companies
Beat the estimates and
you are king …
Miss your estimates and
you get hammered …
For 40 Minutes!
Research in Progress
© Copyright 2008 Dow Jones and Company
|
News Impacts: Some Examples
Rapidly changing market conditions and news events alter
market prices and expectations
Belief is that prices should fully reflect expectations
News events do have significant effects on market prices and
returns
Merger announcements
Earnings guidance
Interest-rate cuts
© Copyright 2008 Dow Jones and Company
|
Bank of America to Buy Countrywide
© Copyright 2008 Dow Jones and Company
|
GE Earnings Guidance
Source: DJ Newswires & MarketWatch
© Copyright 2008 Dow Jones and Company
|
Bank of England Cuts Rates
* DJ Bank of England Cuts Bank Rate To 5.25% From 5.5%
© Copyright 2008 Dow Jones and Company
|
18
Good News/Bad News – Profit from Both
The Ultimate Challenge: Deriving tradable meaning from
the news
Start with publicly available content like DJ News, press releases,
EDGAR filings
Determine analytical methodology
Back-test and determine viability
Integrate multiple models, sources
Apply to real-world
Requirements
Large archive
Text mining tools
Sentiment models
Skilled analysts
© Copyright 2008 Dow Jones and Company
|
19
Potential Alphas
Profiting from expected events
News and momentum:
Evidence that news gives rise to long- and short-term momentum in
prices
Overreaction and underreaction:
Economic data changes
On news releases, stocks overshoot with prices misaligned with
fundamentals
Sentiment - Count of negative news:
A simple count of negative words in news has proven to provide
strategies with positive Alpha
© Copyright 2008 Dow Jones and Company
|
Profiting From News Has Its Challenges
What news events to capture
Economic data, Earnings, Guidance – the easy stuff
Mergers & acquisitions, executive changes, bankruptcies – that’s tough
Sentiment - ??
Many news feeds are not tagged or machine-readable
Need to parse, automate, and digest news
Web-based content is another major challenge
Need to determine:
Which news is relevant? So many sources!
How to react to unexpected events?
Trusted sources
Is news already factored into price?
Which direction will news push the price?
© Copyright 2008 Dow Jones and Company
|
Final Thoughts
Electronic and Quantitative Trading market is rapidly evolving
News is an increasingly important driver of automated trading
strategies
Growing number of tradable news events
News data algorithms
Economic data; earnings and guidance
Elementized News events
Ratings changes, mergers & acquisitions, executive changes
Trading on planned events is profitable
If you can react quickly enough
Not paying attention to news can cost you
Lost Alpha opportunities
Outright losses
© Copyright 2008 Dow Jones and Company
|
Dow Jones Solutions
For Algorithmic & Quantitative Trading
Transforming News For Quantitative Trading And Research
Suite of advanced information and technology tools for a performance edge
Unique news data, tagged and elementized
Computer-readable
Flexible, ultra-low-latency delivery options and seamless integration
News sentiment tools
20-plus-year news archives for back-testing ideas
Tools for creating, deploying and adapting complex, news-based trading strategies
Products
Dow Jones Elementized News Feed
Economic Data – Major economic indicators
Corporate Data – Earnings, Guidance, Mergers, Executive Changes, Ratings,
Bankruptcies, more
Dow Jones News & Archives for Algorithmic Applications
20-plus-year archive of news for analysis
Viewer and toolkit available
Dow Jones News Analytics
News transformed into actionable data
Derive analytics based on sentiment, volume, more
© Copyright 2008 Dow Jones and Company
|
23
Trading Off The News
Questions?
© Copyright 2008 Dow Jones and Company
|
Trading Off The News:
Applications Of News Algorithms
Alan Slomowitz: [email protected]
(201) 938-2195
© Copyright 2008 Dow Jones and Company
|