Prediction Markets
                                  as a Tool of
                                  Crowdsourcing and
                                  Collective Intelligence

                                  Anton Kondratyuk,
Knowledge | Innovation | Talent   Witology
Brief information

Kondratyuk Anton,
Witology

Mathematics and Computer Science, B. A., MSU
Financial Markets and Investment, M. A., HSE

4 years experience in crowdsourcing, idea management and
collective intelligence technologies




                                            Witology | All rights reserved.   2
What is information aggregation?

Information aggregation means gathering some kind of information from the
sources that contain it with help of specific methods and tools.

Crowdsourcing means using wide range of people (community) instead of
experts/specialists for solving some specific task.

                                      Solution


                                                 Information from
                                                 the crowd


                                                            People

                                                               Witology | All rights reserved.   3
Principles of crowdsourcing

• Decentralization (there’s no one source of information for participants)

• Independence (people don’t influence each other)

• Knowledge (participants are not specialists but they must have opinion about
  what they’re asked)

• Diversity (participants must represent various points of view on a problem)

• Motivation (to be sure they give all information they possess)


                                                             Witology | All rights reserved.   4
Effects of crowdsourcing

1. The more people take part, the better results we have.

2. We don’t look for experts, we look for missed “chains” in a crowd
   (in information we get).

3. The efficiency of average participant is not high BUT the overall efficiency
   is high. And every participant is much cheaper than any expert.

4. It’s difficult to find really unique solutions inside group of experts who have
   definite view on the problem field.




                                                                  Witology | All rights reserved.   5
Prediction Markets
emulation of market activity, where future news are traded as assets.




                                                       Witology | All rights reserved.   6
Prediction Markets
aggregate information in prices
exactly as stock markets do

• It’s an instrument of future events prediction
• Traders buy and sell stocks in outcomes of a definite event and in such a way
  push the price up/bring it down.
• When a trader believes in some outcome of the given event, he buys; otherwise
  he sells. The amount of stock he chooses for a transaction depends on his
  confidence in the outcome.
• So if there’s enough liquidity on the market current price is supposed to be an
  effective determinant of beliefs in this outcome of the event, weighted by
  amounts bought&sold.
• BUT unlike stock exchange the price on prediction markets usually fluctuates
  between 0 and 100 and is interpreted as probability measured by traders.
• Traders do what they always do – try to make successful deals according to
  information they have. Probability of the outcome is a “side” product of their
  activity but still the most valuable.
                                                              Witology | All rights reserved.   7
Prediction Markets scheme

                           Multi-outcome                                            Binomial


                                 Barack Obama      60,7%


                                                                                          Yes          60,7%
  Who will win                                                    Will Obama win
  presidential                    Mitt Romney      38,1%   100%     presidential                                 100%
elections 2012?                                                   elections 2012?

                                                                                           No          38,1%


                                 Other candidate   1,9%




      Data source: Intrade.com




                                                                                    Witology | All rights reserved.   8
Use and benefits of Prediction Markets
for society, economics and business

• Prediction Markets are used in all fields which are related to risk management and
  prediction-based decisions such as politics, economics, science, business and corporate
  finance
• Predictions Markets make it possible to aggregate all the information possible at the
  moment
• Market mechanism makes it possible to change prediction by any trader at any time
  according to information change. So it’s different from betting, polls, etc.
• Prediction Markets give the whole picture of the probabilities of outcomes and, what is even
  better, the history of opinions change (information change)
• Prediction Markets are stable to manipulation
• It’s possible to draw tacit and inside information
• The accuracy of Prediction Markets is proved to be better than other methods in not all but
  in many cases.
• Effectiveness of virtual currency PM and real-money PM are similar
• It’s a new way for measuring key business points by its own employeesWitology | All rights reserved.
                                                                                                     9
Open questions about Prediction Markets (PM)

• What scope they can be efficiently applied to?
• There’s a problem of liquidity similar to real markets, so on small unpopular
  markets probabilities can show nothing
• Law hindrances because in many countries PM are treated as gambling (in
  the USA the dispute is still on about permission on public real-money PM)
• The accuracy of PM is worse for low (0-5%) and high (95-100%) probabilities
• There’re some markets which can cause public protests (for example, US
  DARPA market on Osama bin Laden to be caught)




                                                             Witology | All rights reserved.   10
Prediction markets examples

Social markets:
• Intrade (real-money)
• iPredict
• Iowa Electronic Markets (real-money with limit)
• CantorExchange

Corporate vendors for PM:
• ConsensusPoint
• InklingMarkets

Companies using PM for business predictions:
• Google
• HP
• GM
• Pfizer
                                                    Witology | All rights reserved.   11
Social Trading

Social trading is the process through which online financial investors rely mostly (or solely) on
user generated financial content gathered from various applications as the major information
source for making financial trading decisions.
Today the main application for social trading is using other investors’ transactions for making
own decision. It’s a way of “following” other traders to make the strategy as a combination of
other peoples strategies.
So the main competence changes: instead of trying to use technical, fundamental and all other
type of information available traders try only to find investors whose style they like and choose
weights for them to use.
There’re social trading networks:
• eToro
• Zecco
• Zulutrade
• Currensee



                                                                           Witology | All rights reserved.   12
Twitter predicts DJ index
Is it really possible?

Some researches are already done which show correlation between mood state in Twitter and Dow
Jones Industrial Average (DJIA). Moreover, some mood states can predict DJIA 1-3 days before.
The most used mood tweet model by now is GPOMS which includes definite mood states such as:
• Sure
• Happy
• Calm
• Vital
• Alert
• Kind
One really useful state from this set is “calm”. Lagged 2-6 days normalized changes in values show
strong prediction power with DJIA, with accuracy as high as 87,6%. Included into DJIA prediction
models this variable can improve them dramatically.
There are papers that research connection between US indexes and sentiment of twitter-based
community. Results prove that common fear, worry, indifference on the given date can predict
indexes change the next date.

                                                                            Witology | All rights reserved.   13
The most famous correlation
DJIA vs normalized “Calm” value




• Period shown is 28 February, 2008 to 3 November. 2008
• Despite high correlation twitter “calm” value can’t predict some important economic
  events such as large bank bailout which still have much influence on the market
• The origin of this correlation is still not clear
                                                                   Witology | All rights reserved.   14
125047, 35/1 1st Brestskaya st.,
               Moscow, Russia
               +7 499 251 6033
               www.witology.com
  Anton Kondratyuk
                                                  © 2012 Witology.
aakondratyuk@gmail.com                            All rights reserved.

Extent3 witology prediction_markets_2012

  • 1.
    Prediction Markets as a Tool of Crowdsourcing and Collective Intelligence Anton Kondratyuk, Knowledge | Innovation | Talent Witology
  • 2.
    Brief information Kondratyuk Anton, Witology Mathematicsand Computer Science, B. A., MSU Financial Markets and Investment, M. A., HSE 4 years experience in crowdsourcing, idea management and collective intelligence technologies Witology | All rights reserved. 2
  • 3.
    What is informationaggregation? Information aggregation means gathering some kind of information from the sources that contain it with help of specific methods and tools. Crowdsourcing means using wide range of people (community) instead of experts/specialists for solving some specific task. Solution Information from the crowd People Witology | All rights reserved. 3
  • 4.
    Principles of crowdsourcing •Decentralization (there’s no one source of information for participants) • Independence (people don’t influence each other) • Knowledge (participants are not specialists but they must have opinion about what they’re asked) • Diversity (participants must represent various points of view on a problem) • Motivation (to be sure they give all information they possess) Witology | All rights reserved. 4
  • 5.
    Effects of crowdsourcing 1.The more people take part, the better results we have. 2. We don’t look for experts, we look for missed “chains” in a crowd (in information we get). 3. The efficiency of average participant is not high BUT the overall efficiency is high. And every participant is much cheaper than any expert. 4. It’s difficult to find really unique solutions inside group of experts who have definite view on the problem field. Witology | All rights reserved. 5
  • 6.
    Prediction Markets emulation ofmarket activity, where future news are traded as assets. Witology | All rights reserved. 6
  • 7.
    Prediction Markets aggregate informationin prices exactly as stock markets do • It’s an instrument of future events prediction • Traders buy and sell stocks in outcomes of a definite event and in such a way push the price up/bring it down. • When a trader believes in some outcome of the given event, he buys; otherwise he sells. The amount of stock he chooses for a transaction depends on his confidence in the outcome. • So if there’s enough liquidity on the market current price is supposed to be an effective determinant of beliefs in this outcome of the event, weighted by amounts bought&sold. • BUT unlike stock exchange the price on prediction markets usually fluctuates between 0 and 100 and is interpreted as probability measured by traders. • Traders do what they always do – try to make successful deals according to information they have. Probability of the outcome is a “side” product of their activity but still the most valuable. Witology | All rights reserved. 7
  • 8.
    Prediction Markets scheme Multi-outcome Binomial Barack Obama 60,7% Yes 60,7% Who will win Will Obama win presidential Mitt Romney 38,1% 100% presidential 100% elections 2012? elections 2012? No 38,1% Other candidate 1,9% Data source: Intrade.com Witology | All rights reserved. 8
  • 9.
    Use and benefitsof Prediction Markets for society, economics and business • Prediction Markets are used in all fields which are related to risk management and prediction-based decisions such as politics, economics, science, business and corporate finance • Predictions Markets make it possible to aggregate all the information possible at the moment • Market mechanism makes it possible to change prediction by any trader at any time according to information change. So it’s different from betting, polls, etc. • Prediction Markets give the whole picture of the probabilities of outcomes and, what is even better, the history of opinions change (information change) • Prediction Markets are stable to manipulation • It’s possible to draw tacit and inside information • The accuracy of Prediction Markets is proved to be better than other methods in not all but in many cases. • Effectiveness of virtual currency PM and real-money PM are similar • It’s a new way for measuring key business points by its own employeesWitology | All rights reserved. 9
  • 10.
    Open questions aboutPrediction Markets (PM) • What scope they can be efficiently applied to? • There’s a problem of liquidity similar to real markets, so on small unpopular markets probabilities can show nothing • Law hindrances because in many countries PM are treated as gambling (in the USA the dispute is still on about permission on public real-money PM) • The accuracy of PM is worse for low (0-5%) and high (95-100%) probabilities • There’re some markets which can cause public protests (for example, US DARPA market on Osama bin Laden to be caught) Witology | All rights reserved. 10
  • 11.
    Prediction markets examples Socialmarkets: • Intrade (real-money) • iPredict • Iowa Electronic Markets (real-money with limit) • CantorExchange Corporate vendors for PM: • ConsensusPoint • InklingMarkets Companies using PM for business predictions: • Google • HP • GM • Pfizer Witology | All rights reserved. 11
  • 12.
    Social Trading Social tradingis the process through which online financial investors rely mostly (or solely) on user generated financial content gathered from various applications as the major information source for making financial trading decisions. Today the main application for social trading is using other investors’ transactions for making own decision. It’s a way of “following” other traders to make the strategy as a combination of other peoples strategies. So the main competence changes: instead of trying to use technical, fundamental and all other type of information available traders try only to find investors whose style they like and choose weights for them to use. There’re social trading networks: • eToro • Zecco • Zulutrade • Currensee Witology | All rights reserved. 12
  • 13.
    Twitter predicts DJindex Is it really possible? Some researches are already done which show correlation between mood state in Twitter and Dow Jones Industrial Average (DJIA). Moreover, some mood states can predict DJIA 1-3 days before. The most used mood tweet model by now is GPOMS which includes definite mood states such as: • Sure • Happy • Calm • Vital • Alert • Kind One really useful state from this set is “calm”. Lagged 2-6 days normalized changes in values show strong prediction power with DJIA, with accuracy as high as 87,6%. Included into DJIA prediction models this variable can improve them dramatically. There are papers that research connection between US indexes and sentiment of twitter-based community. Results prove that common fear, worry, indifference on the given date can predict indexes change the next date. Witology | All rights reserved. 13
  • 14.
    The most famouscorrelation DJIA vs normalized “Calm” value • Period shown is 28 February, 2008 to 3 November. 2008 • Despite high correlation twitter “calm” value can’t predict some important economic events such as large bank bailout which still have much influence on the market • The origin of this correlation is still not clear Witology | All rights reserved. 14
  • 15.
    125047, 35/1 1stBrestskaya st., Moscow, Russia +7 499 251 6033 www.witology.com Anton Kondratyuk © 2012 Witology. aakondratyuk@gmail.com All rights reserved.