Positive sentiment: trading the news | Financial Industry …
Louis Scott, Founder, Kiema Advisors
“News really does matter but it matters in terms of the volatility it creates rather than the alpha that you can harvest.”
London – The idea that news flow can be
systematically traded is becoming widely accepted. And not just
for major news agencies commenting on planned announcements.
Increasingly, social media is coming to the fore.
Last week, UNICOM’s
fourth annual conference looked at the application of
sentiment analysis on models of trading, fund management and risk
control from a wide range of disciplinary perspectives: pattern
recognition, machine learning, syntax technologies, quantitative
finance, and academia.
Not surprisingly, Thomson Reuters and Bloomberg have both upped
their game in combining financial and editorial aspects of the
business to provide analyses engines on trading screens.
Interest in developing technology is clear – a tweet from the
right account can mean market movements. Recall the “hash crash”
event last year, when a fake tweet from a hacked Associated Press
account reported that US president Barack Obama was injured in a
White House bombing caused the Dow to drop 145 points.
So, there is little doubt social media is being monitored and
measured. But as machine learning gets deployed to make rapid
assessments, the big question is: does sentiment analysis work?
In terms of finding alpha, the jury is out, said Louis Scott,
founder of Kiema Advisors, who was presenting joint research with
Northfield Information Services. A few of the presentations, he
pointed out, throw a wrench into the news analysis engine.
Marco Dion, head of central risk book trading desk at JP
Morgan and previously global head of the firm’s quant research
team, presented research that showed results in terms of alpha
did not translate to huge profitability. Most alpha was found in
small-cap stocks, and by the time transaction costs are added any
value quickly disappeared.
Why alpha disappears was explained by research focused on risk
transfer in the context of informed trades from Anna Obizhaeva,
faculty at New Economic School in Moscow. Her team found that the
more news there is, the more time compresses. The amount of time
for an actionable news event to receive validating information
flow varies widely.
For large caps that time is measured in milliseconds. For small
caps, that could be minutes or even days, said Northfield’s Louis
Scott.
Scott’s presentation went further than the notion that alpha
disappears, but importantly, that risk lingers. He measured
volatility on 400 large caps traded on the NYSE and NASDAQ,
including ETFs (Exchange Traded Funds) and ADRs (American
Depositary Receipts), where news should be most prevalent,
controlling for unanticipated news such as conflicts breaking out
or a CEO being fired.
Scott’s findings show that volatility over a five-day horizon can
be double what it would have been without the news.
“News really does matter but it matters in terms of the
volatility it creates rather than the alpha that you can
harvest,” he said, speaking to Automated Trader. “Machine
learning and applying statistical techniques are suitable for
repeatable experiments or applied to physical systems, which are
predictable, where markets are not.”
Still, the use and analysis of both structured and unstructured
news flow is sure to continue to receive attention and
investment, to some degree because of the implications for market
surveillance and operational risk control.
During a panel discussion, Richard Peterson, CEO of MarketPsych
Data, said that especially in social media, his company
identifies sophisticated attacks to manipulate trading behaviour.
Botnets are hacking accounts on message boards and setting up
Twitter accounts, all the time cultivating a reputation with real
users by, for example, posting generic comments about a company.
A year later, the Botnet recommends buying a stock.
Peterson took his findings to FINRA (Financial Industry
Regulatory Authority), but said there was little interest in
investigating the behaviour. “Basically, they don’t have the
resources to look into it,” he said.
Also speaking on the panel, Simon Townsend, R&D director of
Cloud Compliance at NICE Actimize, said that rolling up news
analytics into trade surveillance adds significant understanding
to what might be behind market movements.
“This really helps us drive up the detection rate for abuse and
most particularly drive down the false positive rate,” he said.
Continue reading:
Positive sentiment: trading the news | Financial Industry …
