Monday, February 6, 2012

Social Wisdom: Prediction Markets

Prediction Markets
Buying and selling a certificate worth a given amount of money if the prediction comes true, but worthless if it doesn't


Putting your 
money where
your mouth is!


People make
predictions
all the time

. . . e.g. about
      other peoples'
      reactions

. . . . . . ability to predict
            the effect of
            your actions on others
            is useful

. . . . . . you need a
           "theory of mind"

. . . . . . you're in a
            game theory
            scenario

. . . . . . data suggests
            competition within groups
            may be a major cause of
            human brain size


Human brain. Credits: Patrick J. Lynch, medical illustrator; C. Carl Jaffe, MD, cardiologist. Http://creativecommons.org/licenses/by/2.5/




Chimpanzee brain. Credits: Gaetan Lee; tilt corrected by Kaldari. Http://creativecommons.org/licenses/by/2.0/



. . . . . . this is the
            social competition theory
            of human brain genesis
           (ref.: Bailey & Geary,
           Hominid Brain Evolution:
           Testing Climatic,
           Ecological, and Social Competition
           Models
           Human Nature,
           vol. 20, no. 1,
           Mar. 2009, pp. 67-79.)


. . . . . . so the need to predict
            may explain
            why we're human!
            (Predict what?)



Prediction is
important in 
other areas too

Consider
the weather:


Credits:
Don Amaro from
Madeira Islands,
Portugal,
upload by Herrick
17:17, 4 December 2007 (UTC).
Http://creativecommons.org/licenses/by/2.0/


Waterspout.
Credits: http://www.photolib.noaa.gov/bigs/wea00308.jpg.
1969 September 10.
Photographer: Dr. Joseph Golden,
NOAA. Public domain.

Two images of Hurricane Irma














. . . The National Weather Service (NWS)
       is a large, highly technical
       gov't agency devoted to
             ~ prediction ~
       using large computers, etc.
. . . It's not perfect
      but it's
      much better than nothing

. . . weather prediction
      uses computers

. . . . . . other kinds of
            prediction can use
            other methods



Crowdsourcing group wisdom...

     Horse race betting
     is an example

Sourcehttps://www.greatfallstribune.com/story/news/2019/12/09/montana-horse-racing-board-pays-628-k-debt-state-early/2630053001

           Bet right, you win $

           Bet wrong, you lose $ 
  
           But if you win, how much $?

                It depends on how much was bet on what horse!

                A crowd-sourced estimate of the odds

                A kind of prediction market

Pari-mutuel betting is typical

Take all $ bet on any horse

Divide it up among bets on the winning horse

From that algorithm you can calculate odds etc.

---

For "place" bets, take all place bets

Divide half among bets on the 1st place horse

Divide other half among bets on the 2nd place horse 

---

For "show" bets, take all place bets

Divide 1/3 among bets on the 1st place horse

Divide 1/3 among bets on the 2nd place horse 

Divide 1/3 among bets on the 3rd place horse 

 

               

           Other sports betting odds are set ahead of time

               Traditionally by bookies in a back room
  
               Experts combining opinions

                   But not using Delphi method

                   Would it work better if they used Delphi method?


Prediction Market

     Stock market:

         Buy if expect it to go up
         Were you right? Make $
         Were you wrong? Lose $

        Stock market is a kind of...

        Prediction Market
       
            Other ways of 
            looking at the
            stock market

                What are they?




     An example from the history of this course

     . . . Will the
     average global temperature
     for 2012
     be the highest 
     ever recorded?


     . . . at 12:23 on the afternoon before class (Sp '12):
     you could have bought a "yes"
     for $2.50 from intrade.com
     . . . if you ended up right -
           you'd get $10
     . . . if wrong -
           you'd lose your $2.50

    What ultimately happened?



















        Jan 16, 2013:

       NASA: 2012 Was 
            9th Hottest Year 
            on Record

As of early 2020:





Let's go to 
intrade.com 
and check it out



Here are/were some other prediction markets:


Ok, here's another problem 
with prediction markets

    People might bet 
    for other reasons 
    than to make $

        Why might someone do that?







Bet on something you don't want to happen

  • If it happens anyway, you'll feel better!


Hedge a financial loss by betting it will happen

  • Stock market might decline 
    • So, buy a bet that it will!
  • Purely personal issues would not work
    • Why not?



Manipulate odds to affect reality

Can you think of 
a possible example?

       Example:
       Manipulate press coverage 
       of a presidential election

             It's been done!


             It's expensive!
             Why do it?


Why should Intrade
prediction contracts
get closer to
$10 or $0,
PredictIt contracts
get closer to $1 or $0,
LR  Prediction Company* contracts
get closer to $100 or $0, 
etc.,
as they approach
resolution?

*???
 
Could price ever get
high, then end up low
(or vice versa)?

We started calling them "bets" ...

        But technically it's a "prediction market contract"

  • It's still a bet!
 


5. More about sports prediction markets

Tim Henman
serving at Wimbledon, 2005.
Credits: Photo by Spiralz, license by http://creativecommons.org/licenses/by/2.0/

. . . Sports betting has
       a long history


 . . . People want to
       predict games
       
       They'll pay
       good money
       to do it!

. . . A theoretically ideal
      "honest bookmaker"
      will not expect to make a profit



. . . . . The $$ bet are 
          merely redistributed


. . . (But how do
      real bookmakers
      differ from that?)

. . . . . This is how horse racing works


6. Political prediction...

. . .During election season,
     media and candidates
     all try to
     predict outcomes


. . . Some of it
      you don't hear about


. . . polling,
      trend analysis,
      sociological analysis
      are big

      . . . but why is
      asking people
      their opinion
      unreliable?


. . . predictions markets
      have been claimed
      to do it better!

. . . But they aren't perfect

  • Manipulation can skew results
  • People can buy predictions "irrationally"
    • Example: 2012 election


7. About corporate stocks...


. . . . . . The stock market
            is a
            "leading economic indicator"


. . . . . . (Economists pay
            special attention to
            leading indicators)

. . . . . . Is the stock market
            a prediction market?



8. If someone
asked you
to invent a way
to collect
group predictive wisdom,
what might you come up with?



. . . Maybe a
      Delphi-like method


. . . Probably (?) not
      prediction markets

. . . . . .However, by accident,
           horse racing is in that direction


. . . . . . An early design for
            prediction markets
            appears in
            The Shockwave Rider,
            by John Brunner, 1975


. . . . . . His term was Delphi Pool


. . . . . . General idea:
            when real money
            is at stake, 
            people predict better

                  Do you agree?



9. Terrorism and
    prediction markets


. . . . . . DARPA's PAM
            (Policy Analysis Market)
            permitted a
            prediction market
            for terrorist attacks

. . . . . . It worked like this:
            How much would you pay
            to get $100 if there is a
            snowstorm tomorrow?
            Trick 1: buy a snowstorm prediction cheap
                 Make a snowstorm happen, then collect $100
            Similarly for a terrorist attack
            Trick 2: watch for terrorism contracts
                 Arrest them first


. . . . . . Does that
            sound like
            a good idea
            to you?

. . . . . . How might it 
            help fight terrorism?










+ we might have forwarning

- terrorists might
  buy predictions, then
  make them come true
  to make $$$!


. . . . . . In 2003,
            2 senators found out,
            PAM was cancelled, and
            a DARPA
            program director resigned


. . . . . . . . . not clear if
                  terrorism predictions
                  were ever traded



10. As time allows, 
let's 

      - try our own prediction market!    

               Prediction market contract:
                       see syllabus

About how to do it, as well as some of the finer points: 
     The rational participant would be willing to buy a contract for a little less than the probability of the event occurring (and thus paying off), and sell for a little more than that probability. For example if the event was a six resulting from a roll of a fair die, it would be worth buying a contract for that for $16 and selling one for $17 for a contract worth $100 (since the probability of the event is 16.67%). 
     To create new contracts, it works for the broker to simply lend out blank contracts for free; if the event later occurs the person taking one would have to then pay the broker $100, and if the event does not occur the contract is worthless and the broker is not paid anything.  
     It is an interesting point that the originator has a different risk
exposure. So they would demand a risk premium, suggesting that the
price should decline a bit over time, not because of any changes in
probabilities in the domain of the contract, but just as an artifact
of the origination risk. As the broker and proprietor of the "Little
Rock Prediction Market Co." I get to make the rules and I made them to
be useful in an in-class simulated prediction market. In a normal
semester we would be buying and selling them every week until the end
of the semester (or the event the contracts are about occurs,
whichever happens first). Then at the end of the semester I'd show a
graph of the price over time, and send everyone an email explaining
how much money they won or lost.
    I could of course make bad rules, and in that case the rational
market player would play to win, leading to prices that do not reflect
the underlying probabilities. 
    A similar origination problem might occur with real prediction
markets - somehow the market must be primed with new contracts, and
how are they to be priced? If there are lots and lots of buyers
compared to the number of contracts, the contracts could just be sold
to the highest bidders. I can't do that in class because there aren't
enough students and also because it would be confusing to many people.
Another problem like this occurs with stock IPOs - the initial price
is just a guess (I think? Or maybe it is manipulated to be too low so
that the first purchasers have a better chance of cleaning up?) and
the price can then swing wildly at first, exposing early buyers to
increased risk. Another situation like that is when a prediction
market reaches its end. If it rapidly becomes clear what the outcome
is, the people still holding the contracts won't be able to buy and
sell them fast enough to track the events on the ground, exposing them
to the same kind of increased risk. In an extreme case, suppose the
contracts were about the outcome of a roll of the dice. The outcome
goes from pure chance to completely settled in an instant, exposing
late buyers to increased risk.
    Another aspect to the market as designed for this course is it
accommodates uncertainty about the probability. For example, you might
not be confident of your estimate of 20% probability, but more
comfortable guessing that it is somewhere between 20% and 50%. Then
you would be willing to buy for $19 and sell for $51. If you were more
risk-tolerant, you might reason that the real probability is more
likely to be somewhere in the middle region than very near 20% or 50%,
so you might then be willing to buy for $25 and sell for $45, if
needed in order to execute a transaction.

Disclaimer: 
Using gen-u-ine, real $10 bills would make this a very memorable learning experience. Thus, it would have educational benefit! Nevertheless, we will only pretend, because this course wishes to avoid a legal minefield of potential complications.


No comments:

Post a Comment