deriving trading signals from google trends and wikipedia page views

32
Deriving Trading Signals from Google Trends and Wikipedia Page Views Thomas Wiecki

Upload: twiecki

Post on 29-Jan-2015

124 views

Category:

Economy & Finance


0 download

DESCRIPTION

This was a talk I gave at the http://www.Quantopian.com meet-up in Boston and NYC.

TRANSCRIPT

Page 1: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Deriving Trading Signals from Google Trends and

Wikipedia Page ViewsThomas Wiecki

Page 2: Deriving Trading Signals from Google Trends and Wikipedia Page Views

This is Joe.He is worried about the debt ceiling.

Page 3: Deriving Trading Signals from Google Trends and Wikipedia Page Views

What does he do?

Page 4: Deriving Trading Signals from Google Trends and Wikipedia Page Views
Page 5: Deriving Trading Signals from Google Trends and Wikipedia Page Views

After gathering information he calls his broker.

Page 6: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Who sells all of his clients stock.

Page 7: Deriving Trading Signals from Google Trends and Wikipedia Page Views
Page 8: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Stock market 101

The price is the result of the trading decisions of many individuals.

Page 9: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Decision Making:Multiple stages

Page 10: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Motivation

Quantify information gathering behavior that precedes investment decisions.

Page 11: Deriving Trading Signals from Google Trends and Wikipedia Page Views

● Stock prices follow news.● News can't be predicted ⇒ Random walk.● However: Stocks do not follow random walk.● What about bubbles?● More and more research casting doubt...

Efficient Market Hypothesis

Page 12: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Quantitative Behavioral Finance

● Online chat activity predicts books sales [1]● Blog sentiment analysis predicts movie sales

[2].● Google search queries predict disease

infection and consumer spending [3].● ⇒ News impact markets, but so does public

mood and sentiment.

Page 13: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Cognitive Bias: Loss Aversion

Page 14: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Subject of recent research

Page 15: Deriving Trading Signals from Google Trends and Wikipedia Page Views
Page 16: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Simple investment strategy based on Google search volumefor t in [1:T]:

avg_search_vol = mean(search_vol[t-2:t-5])

if search_vol[t-1] > avg_search_vol:

short DJIA for one week

if search_vol[t-1] < avg_search_vol:

long DJIA for one week

Page 17: Deriving Trading Signals from Google Trends and Wikipedia Page Views
Page 19: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Top predictors

Page 20: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Bottom predictors

Page 21: Deriving Trading Signals from Google Trends and Wikipedia Page Views
Page 22: Deriving Trading Signals from Google Trends and Wikipedia Page Views
Page 23: Deriving Trading Signals from Google Trends and Wikipedia Page Views
Page 24: Deriving Trading Signals from Google Trends and Wikipedia Page Views
Page 26: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Twitter Sentiment Analysis

Page 27: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Can Twitter move the market?

Page 28: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Cautionary Tale

Page 29: Deriving Trading Signals from Google Trends and Wikipedia Page Views

● Founded February 2011● Closed after one month in service...● However: return of 1.86% (beating the

market and average hedge fund)

Twitter Fund (Derwent Capital Markets)

Page 30: Deriving Trading Signals from Google Trends and Wikipedia Page Views

● Preliminary evidence that information gathering can be quantified and exploited.

● Quantopian - Reproducibility Science● Mountains of data, waiting to be explored!

Departing thoughts...

Page 31: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Thanks! Questions?

Contact:● [email protected]● Twitter: @twiecki● GitHub: twiecki

Page 32: Deriving Trading Signals from Google Trends and Wikipedia Page Views

Image sources and references● http://www.ng.all.biz/img/ng/service_catalog/502.jpeg● http://www.123rf.com/photo_10037927_businessman-or-stock-broker-with-cellphone.html● http://www.financetwitter.com/wp-content/uploads/2011/08/SP500_Crash_4Aug2011.jpg● http://lydiakimblesellsvegas.com/images/buy-sell-keyboard.jpg● http://venturebeat.com/2012/05/28/twitter-fueled-hedge-fund-bit-the-dust-but-it-actually-worked/● Gilbert, E & Karahalios, K. (2010) Widespread worry and the stock market.● [11] Gruhl, D, Guha, R, Kumar, R, Novak, J, & Tomkins, A. (2005) The predictive power of online

chatter. (ACM, New York, NY, USA), pp. 78–87.● Mishne, G & Glance, N. (2006) Predicting Movie Sales from Blogger Sentiment. AAAI 2006

Spring Symposium on Computational Approaches to Analysing Weblogs● S. Asur and B. A. Huberman 2010 Predicting the Future with Social Media arXiv:1003.5699v1● Choi, H & Varian, H. (2009) Predicting the present with google trends., (Google), Technical

report.● Liu, Y, Huang, X, An, A, & Yu, X. (2007) ARSA: a sentiment-aware model for predicting sales

performance using blogs. (ACM, New York, NY, USA), pp. 607–614.