prediction, belief, and markets -...
Transcript of prediction, belief, and markets -...
![Page 1: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/1.jpg)
Prediction, Belief, and Markets
http://aaaimarketstutorial.pbworks.com
Jake Abernethy, UPenn è UMich Jenn Wortman Vaughan, MSR NYC
July 14, 2013
![Page 2: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/2.jpg)
Belief, Prediction, and… Gambling? A Short History Lesson
• The earliest references to probability calculations arose directly through the study of games of chance, like dice.
• We have evidence of human gambling as early as ancient Egypt, more than 3000 years ago.
• We have no evidence of probabilistic analysis until the time Pascal in the late 1600s.
![Page 3: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/3.jpg)
The Emergence of Probability
![Page 4: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/4.jpg)
Probability and Finance: It’s Only a Game!
![Page 5: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/5.jpg)
Letters b/t Pascal and Fermat, 1654 • The beginning of mathematical probability is often dated
to letters between Pierre de Fermat and Blaise Pascal. • The setting discussed by Pierre and Fermat: two players
are playing a game, they need to get n points to win, and the winner receives prize.
• How to divide the prize if the game is cut short?
![Page 6: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/6.jpg)
Bruno de Finetti: “PROBABILITY DOES NOT EXIST”
• De Finetti, well-known in probability theory, had the view that we may only think about probabilities in terms of rates of betting.
• That is, the laws of probability can be viewed as resulting from simple “no-arbitrage” conditions on these rates
![Page 7: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/7.jpg)
Betting can be used to elicit beliefs • Economists, in particular, are very keen on betting with
each other. Alex Tabarrok, in marginalrevolution.com: “A Bet is a Tax on Bullshit”
• There’s been some recent debate in the Econ blogosphere about the extent to which bets really do reveal beliefs. Noah Smith, of noahpinion.com: “The mistake is looking at the risk and return of single assets instead of portfolios. Basically, the risk of an asset… is based mainly on how that asset related to other assets in your portfolio.”
![Page 8: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/8.jpg)
How do I find someone to bet with? Answer: A Prediction Market
• Prediction markets have existed for over 200 years. • Typically, odds (prices) are set by supply and demand. • People began to notice: the market prices are generally
very accurate, and provide better predictors than expert assessments, etc. What’s going on?
• Robin Hanson: “Rational expectations theory predicts that, in equilibrium, asset prices will reflect all of the information held by market participants. This theorized information aggregation property of prices has lead economists to become increasingly interested in using securities markets to predict future events.”
![Page 9: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/9.jpg)
Outline Before the break: 1. Predictions Markets in Practice 2. Eliciting beliefs with proper scoring rules 3. Bregman divergences + proper scoring rules 4. Hanson’s Market Scoring Rule
After the break: 1. Securities markets 2. Duality & connections to online learning 3. Handling very large outcome spaces 4. Overview of additional topics
![Page 10: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/10.jpg)
The Most (in)Famous Prediction Market
• 1999: Intrade founded by John Delaney • 2003: Acquired by TradeSports in 2003, later splits off
after TradeSports closes down in 2008 • 2004: Intrade gains notoriety during Bush/Kerry election
for providing continuous forecasts throughout campaign • May 2011: Founder John Delaney dies at age 42 while
climbing Mt. Everest, less than 50 meters from summit • Nov. 2012: US regulator CFTC files suit against Intrade,
leading Intrade to disallow US customers from betting • Mar. 2013: Due to “financial irregularities”, Intrade halts
trading, freezes all accounts. Still remains in legal limbo.
![Page 11: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/11.jpg)
Example: Intrade
![Page 12: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/12.jpg)
Iowa Electronic Markets (IEM): Legal and with Real Money
• Founded in 1988 at the University of Iowa for the purpose of research in market prediction accuracy
• Received a “no action” letter from the CFTC, permitting them to facilitate unregulated betting. (Such letters are apparently “no longer being given out”)
• On the downside, the IEM must obey a certain set of conditions. Most notably, individual traders may deposit no more than $500.
![Page 13: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/13.jpg)
Predictious: A New Bitcoin-based Prediction Market
![Page 14: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/14.jpg)
Example: Inkling Markets
Internal prediction markets used within companies
![Page 15: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/15.jpg)
Markets in Practice Questions: 1. What are different market “mechanisms”? 2. How quickly do markets incorporate
information? 3. How accurate are market prices, vis-à-vis
prediction?
![Page 16: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/16.jpg)
“Arrow-Debreu Security”: Contract pays $10 if X happens, $0 otherwise. If I think that Pr(X) = p then I should: • Buy this security at any price less than $10p • Sell this security at any price greater than $10p
Current price measures the population’s collective beliefs
“Arrow-Debreu” Securities
![Page 17: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/17.jpg)
[1] Market Mechanisms: Continuous Double Auction (CDA)
• Used by Intrade.com and Betfair.com
• Market receives a sequence of orders
• Two types of orders: • Limit order: trader
posts shares to order book
• Market order: trader buys shares in order book
![Page 18: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/18.jpg)
Obama2012 Intrade: Bid+Ask+Trades
![Page 19: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/19.jpg)
Aside: Problems with the CDA • Chicken and egg problem: who is willing to join a market
if there are no other participants? • Not a lot of “liquidty”: it’s very easy to swing prices • Large bid/ask spreads
• Alternative mechanism: the automated market maker, which we will be discussing throughout the 2nd half of the tutorial
![Page 20: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/20.jpg)
[2] How Quickly do Markets Respond?
Source: Snowberg, Wolfers, Zitzewitz 2012
![Page 21: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/21.jpg)
[3] Are Market Prices Accurate? • The market price for Arrow-Debreu security is essentially
a “consensus estimate” of the probability of an event
• Are these estimates accurate?
• We can check this on historical data, but… • Prices are changing, which price do we use? • What is the right metric to measure accuracy? • What are we comparing against?
![Page 22: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/22.jpg)
Market Prediction vs. True Vote Share
Berg et al., 2008: “Results From a Dozen Years of Election Futures Markets Research”
![Page 23: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/23.jpg)
Average Polls vs. Market Prices
Poll Error: (average from last week)
1.91%
Market Error: (election eve)
1.49%
Market Error: (average from last week)
1.58%
Berg et al., 2008: “Results From a Dozen Years of Election Futures Markets Research”
![Page 24: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/24.jpg)
Aside: Supreme Court + Health Care Intrade market: “The US Supreme Court to rule individual
mandate unconstitutional before midnight ET 31 Dec 2012”
![Page 25: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/25.jpg)
More on Obamacare Prediction Market • David Leonhardt in the NYTimes: “After several years in
which the market was often celebrated as a crystal ball, the Supreme Court ruling was a useful corrective. The prediction-market revolution, like so many others, initially promised more than it could deliver.”
• Response by Robin Hanson on overcomingbias.com: “But the Intrade market on the Obamacare court case was an active valid market, on an appropriate subject. When it assigned a 75% chance to an event it was saying real loud that it would be wrong 1/4 of the time. And studies have consistently found such markets are well-calibrated in this way. What more do you want?”
![Page 26: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/26.jpg)
The Basics: Proper Scoring Rules
![Page 27: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/27.jpg)
1950: Brier on Weather Forecasting
![Page 28: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/28.jpg)
![Page 29: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/29.jpg)
![Page 30: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/30.jpg)
How Should We Pay a Forecaster?
What is the “right” payment scheme to reward/punish a forecaster who makes a sequence of probability predictions for events that we observe?
• The sequence of outcomes:
• The sequence of forecasts:
• The forecaster’s payment:
y1, y2, y3, ...∈ {0,1}
p1, p2, p3, ...∈ [0,1]
€
1T
S(yt , pt )t=1
T
∑
![Page 31: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/31.jpg)
Brier Score ó Quadratic Scoring Rule
• For a binary outcome y ∈ {0,1}, p ∈ [0,1]
• For one of n outcomes, y ∈{1,...,n}, p ∈ ∆n
S(y, p) = −(y− p)2
S(y,p) = − (1y=i − pi )2
i=1
n
∑
![Page 32: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/32.jpg)
What’s Special About This Function?
Assume y is random and Pr(y = 1) = q. Then...
€
p∈[0,1]argmax Ε −(y − p)2[ ]( )
S(y, p) = −(y− p)2
€
=p∈[0,1]argmax −(p − q)2 − q + q2( ) = q
€
=p∈[0,1]argmax −q(1− p)2 − (1− q)p2( )
![Page 33: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/33.jpg)
Proper Scoring Rules • What we have just introduced is the notion of a proper
scoring rule, any function S satisfying
• The scoring rule is said to be strictly proper if the above inequality is strict unless p = q
€
Εy~q S(y,q)[ ] ≥ Εy~q S(y,p)[ ] ∀p,q∈ Δ n
![Page 34: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/34.jpg)
Another Strictly Proper Scoring Rule
• This is known as the logarithmic scoring rule. For binary random variables, it can be written as:
• EXERCISE: check that this is proper!
€
S(y,p) = log p(y)
S(y, p) =log p y =1
log(1− p) y = 0
"#$
%$
![Page 35: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/35.jpg)
• Effectively, a scoring rule is just a type of loss function
• Scoring rules measure the performance (not loss) of a predicted distribution given a final outcome
• Research on scoring rules is focused more heavily on the
incentives of the associated payment mechanism
Scoring Rules == –Loss Functions?
![Page 36: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/36.jpg)
Designing Scoring Rules Using Bregman Divergences
![Page 37: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/37.jpg)
Savage 1973
![Page 38: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/38.jpg)
Digression: Bregman Divergences • A Bregman divergence measures distance with respect to
a convex function f
€
Df (p,q) = f (p) − f (q) −∇f (q) ⋅ (p − q)
![Page 39: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/39.jpg)
Digression: Bregman Divergences
• Properties:
€
Df (p,q) ≥ 0 ∀p,q
€
Df (p,q) ≠ Df (q, p) (in general)€
Df (p, p) = 0 ∀p€
Df (p,q) = f (p) − f (q) −∇f (q) ⋅ (p − q)
![Page 40: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/40.jpg)
Bregman Divergences III
• Example 1, quadratic:
• Example 2, entropic:
€
f (p) = || p ||2 ⇒ Df (p,q) = || p − q ||2
f (p) = pi log pii∑ ⇒ Df (p,q) = pi log
piqii
∑
€
Df (p,q) = f (p) − f (q) −∇f (q) ⋅ (p − q)
![Page 41: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/41.jpg)
Bregman Diverg.ó Scoring Rule?? • Let f be any convex function • Let ei be the ith indicator vector, ei = 〈0,…,0,1,0,…,0〉 • Let p, q be any two distributions • Fact: There exists a function g such that
and so
• This is the scoring rule property!!
€
E i~q Df (e i,p)[ ] = Df (q,p) + g(q)
€
argmaxp∈Δn
E i~q −Df (e i,p)[ ] = q
![Page 42: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/42.jpg)
Bregman Diverg.ó Scoring Rule!! • We now have a recipe for constructing scoring rules:
Take any convex function f and set
• Quadratic Scoring Rule:
• Log Scoring Rule: €
S(i,p) = −Df (e i,p)
f (p) =||p ||22
f (p) = pi log pii∑
![Page 43: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/43.jpg)
Brief Literature Review
![Page 44: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/44.jpg)
Market Scoring Rules for Belief Aggregation
![Page 45: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/45.jpg)
Learning a Consensus? • Scoring rules are useful for incentivizing one individual to
state his beliefs about a probability, but what if we’d like to learn from a crowd
• Proposal: We could just pay every individual according to a scoring rule.
• Problems: • This could be very expensive! • How should we combine estimates? • How can we weed out noise traders? • How can we weed out copycats?
![Page 46: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/46.jpg)
Market Scoring Rules Robin Hanson proposed the following idea to create a prediction market based on an automated market maker:
• Suppose we have a random variable X which will take one of n values {1, 2, …, n}
• The MM chooses a scoring rule S and announces it • The MM then posts an initial distribution (prior) p0
• Traders arrive, one-by-one, giving updates pt-1 → pt
• Eventually, outcome X is revealed, and trader t earns (or loses)
S(X,pt )− S(X,pt−1)
![Page 47: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/47.jpg)
Market Scoring Rule
B C
A
p0
pt p2
p1
![Page 48: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/48.jpg)
Incentives and Costs • Assume trader t has belief distribution p on X, which can
(and should!) depend on previous market observations • Suppose he wants to maximize his payment
• The MM must make all payments, which total
• This is bounded! This is like MM’s subsidy to market.
argmaxpt ΕX~p[S(X,pt )− S(X,pt−1)]
€
S(X,pt ) − S(X,pt−1)[ ]t=1
T
∑ = S(X,pT ) − S(X,p0)
= p= argmaxpt ΕX~p[S(X,pt )]always non-negative!
![Page 49: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/49.jpg)
LMSR: Log Market Scoring Rule • Initial hypothesis p0 is the uniform distribution • Trader t posts an update pt-1 → pt • After X is revealed, trader t earns log(pt(X)/pt-1(X))
• Hanson: the LMSR is an important special case, the only MSR for which “betting on conditional probabilities does not affect marginal probabilities”
• The market maker’s worst case loss is bounded by log n, where n is the number of possible values of X
![Page 50: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/50.jpg)
Prediction, Belief, and Markets: Part 2
Jake Abernethy, UPenn → UMich Jenn Wortman Vaughan, Microsoft Research NYC
http://aaaimarketstutorial.pbworks.com
![Page 51: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/51.jpg)
Outline of Part 2
1. Automated market makers for securities markets 2. Designing markets for large outcome spaces 3. Duality & connections to online learning 4. Recent extensions and additional topics
![Page 52: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/52.jpg)
Back to Securities Markets
![Page 53: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/53.jpg)
Arrow–Debreu Securities
![Page 54: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/54.jpg)
Potential payoff is $10. If I think that the probability of this event is p, I should • Buy this security at any price less than $10p • Sell this security at any price greater than $10p
Current price measures the population’s collective beliefs
Arrow–Debreu Securities
![Page 55: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/55.jpg)
How do we arrive at the current price?
Arrow–Debreu Securities
![Page 56: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/56.jpg)
How do we arrive at the current price? • Traditional stock market style pricing (continuous double
auction) – low liquidity, huge spreads
Arrow–Debreu Securities
![Page 57: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/57.jpg)
How do we arrive at the current price? • Traditional stock market style pricing (continuous double
auction) – low liquidity, huge spreads • Automated market maker – willing to risk a (bounded)
loss in order to encourage trades
Arrow–Debreu Securities
![Page 58: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/58.jpg)
• In a complete market, a security is offered for each of a set of mutually exclusive and exhaustive events
Market Makers for Complete Markets
![Page 59: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/59.jpg)
• In a complete market, a security is offered for each of a set of mutually exclusive and exhaustive events
Market Makers for Complete Markets
Worth $1 iff
Worth $1 iff
![Page 60: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/60.jpg)
• In a complete market, a security is offered for each of a set of mutually exclusive and exhaustive events
• An automated market maker is always willing to buy and sell these securities at some price
Market Makers for Complete Markets
Worth $1 iff
Worth $1 iff
![Page 61: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/61.jpg)
Cost Functions
![Page 62: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/62.jpg)
Already purchased: q1 shares q2 shares
Cost Functions
Worth $1 iff
Worth $1 iff
![Page 63: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/63.jpg)
Already purchased: q1 shares q2 shares Want to purchase: r1 shares r2 shares
Cost Functions
Worth $1 iff
Worth $1 iff
![Page 64: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/64.jpg)
Cost of purchase: C(q + r) – C(q)
Already purchased: q1 shares q2 shares Want to purchase: r1 shares r2 shares
Cost Functions
Worth $1 iff
Worth $1 iff
![Page 65: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/65.jpg)
Cost of purchase: C(q + r) – C(q)
Already purchased: q1 shares q2 shares Want to purchase: r1 shares r2 shares
Instantaneous prices: p1 = ∂C / ∂q1 p2 = ∂C / ∂q2
Cost Functions
Worth $1 iff
Worth $1 iff
![Page 66: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/66.jpg)
Cost of purchase: C(q + r) – C(q)
Already purchased: q1 shares q2 shares Want to purchase: r1 shares r2 shares
Instantaneous prices: p1 = ∂C / ∂q1 p2 = ∂C / ∂q2
Cost Functions
Worth $1 iff
Worth $1 iff
“predictions”
![Page 67: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/67.jpg)
Back to the LMSR Remember the logarithmic market scoring rule…
• Initial hypothesis p0 is the uniform distribution
• Trader t posts an update pt-1 → pt
• After outcome i is revealed, trader t receives log(pt,i) – log(pt-1,i) = log(pt,i / pt-1,i)
![Page 68: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/68.jpg)
The logarithmic market scoring rule can be implemented as a cost function based market with cost function
and instantaneous prices
Back to the LMSR
pi = exp(qi ) Σj exp(qj )
€
C(q1,...,qN) = log exp(qi)i = 1
N∑
![Page 69: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/69.jpg)
The logarithmic market scoring rule can be implemented as a cost function based market with cost function
and instantaneous prices Notice that pi is increasing in qi and the prices sum to 1
Back to the LMSR
pi = exp(qi ) Σj exp(qj )
€
C(q1,...,qN) = log exp(qi)i = 1
N∑
![Page 70: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/70.jpg)
Equivalence
For all p, p', q, q', such that ∇C(q) = p and ∇C(q') = p', for all outcomes i, a trader who changed the market state from p to p' in the MSR would receive the same total payoff as a trader who changed the market state from q to q' in the cost function based market.
[Hanson 03; Chen & Pennock 07]
![Page 71: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/71.jpg)
A Proof in One Slide
![Page 72: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/72.jpg)
A Proof in One Slide
€
= q'i −qi( ) − C(q') −C(q)( )
cost function payoff
security value when the outcome i occurs
cost of the purchase
![Page 73: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/73.jpg)
A Proof in One Slide
€
= q'i −qi( ) − C(q') −C(q)( )
cost function payoff
![Page 74: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/74.jpg)
A Proof in One Slide
€
= q'i −qi( ) − log eq' jj∑ − log eq j
j∑
$
% & &
'
( ) )
€
= q'i −qi( ) − C(q') −C(q)( )
cost function payoff
by definition
![Page 75: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/75.jpg)
A Proof in One Slide
€
= q'i −qi( ) − log eq' jj∑ − log eq j
j∑
$
% & &
'
( ) )
€
= q'i −qi( ) − C(q') −C(q)( )
cost function payoff
![Page 76: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/76.jpg)
A Proof in One Slide cost function payoff
€
= q'i −qi( ) − C(q') −C(q)( )
€
= logeq' i − logeqi( ) − log eq ' jj∑ − log eq j
j∑
$
% & &
'
( ) )
![Page 77: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/77.jpg)
A Proof in One Slide cost function payoff
€
= q'i −qi( ) − C(q') −C(q)( )
€
= logeq' i − logeqi( ) − log eq ' jj∑ − log eq j
j∑
$
% & &
'
( ) )
€
= logeq ' i
eq ' jj∑
− logeqi
eq j
j∑
![Page 78: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/78.jpg)
A Proof in One Slide cost function payoff
€
= q'i −qi( ) − C(q') −C(q)( )
€
= logeq' i − logeqi( ) − log eq ' jj∑ − log eq j
j∑
$
% & &
'
( ) )
€
= logeq ' i
eq ' jj∑
− logeqi
eq j
j∑ prices!
![Page 79: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/79.jpg)
A Proof in One Slide cost function payoff
€
= q'i −qi( ) − C(q') −C(q)( )
€
= log p'i −log pi€
= logeq' i − logeqi( ) − log eq ' jj∑ − log eq j
j∑
$
% & &
'
( ) )
€
= logeq ' i
eq ' jj∑
− logeqi
eq j
j∑
![Page 80: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/80.jpg)
A Proof in One Slide cost function payoff
€
= q'i −qi( ) − C(q') −C(q)( )
€
= logeq ' i
eq ' jj∑
− logeqi
eq j
j∑
= scoring rule payoff
€
= log p'i −log pi€
= logeq' i − logeqi( ) − log eq ' jj∑ − log eq j
j∑
$
% & &
'
( ) )
![Page 81: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/81.jpg)
More Generally
• Any market scoring rule can be implemented as a cost function based market [Chen & Pennock 07; Chen & Vaughan 10; Abernethy & Frongillo 11; …]
![Page 82: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/82.jpg)
More Generally
• Any market scoring rule can be implemented as a cost function based market [Chen & Pennock 07; Chen & Vaughan 10; Abernethy & Frongillo 11; …]
• Advantages: • Retains the good incentive properties of the MSR • Arguably more “natural” for traders • Exposure to risk is more transparent
![Page 83: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/83.jpg)
Beyond Complete Markets
![Page 84: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/84.jpg)
Complex Outcome Spaces
n! 2n Infinite
![Page 85: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/85.jpg)
Complex Outcome Spaces
n! 2n Infinite
MSR-NYC’s WiseQ [Dudik et al., 2013]
![Page 86: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/86.jpg)
• Cannot simply run a standard market like LMSR • Calculating prices is intractable [Chen et al., 2008] • Reasoning about probabilities is too hard for traders
Complex Outcome Spaces
n! 2n Infinite
![Page 87: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/87.jpg)
• Cannot simply run a standard market like LMSR • Calculating prices is intractable [Chen et al., 2008] • Reasoning about probabilities is too hard for traders
• Can run separate, independent markets (e.g., horses to win, place, or show) but this ignores logical dependences
Complex Outcome Spaces
n! 2n Infinite
![Page 88: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/88.jpg)
Complex Outcome Spaces
Given a small set of securities over a very large (or infinite) state space, how can we design a consistent market that can be operated efficiently?
[Abernethy, Chen, and Vaughan, EC 2011; long version in ACM TEAC 2013]
![Page 89: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/89.jpg)
Menu of Securities We would like to offer a menu of securities{1, …, K}
specified by a payoff function ρ
![Page 90: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/90.jpg)
Menu of Securities We would like to offer a menu of securities{1, …, K}
specified by a payoff function ρ
10 0 5.5 0 17 0 .9 .9 .9 0 .9 .9 0 42 0 10 10 10 0 0 11.5 8 0 0 1 0 0 0 0 1
securities
outcomes
payoff
![Page 91: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/91.jpg)
Example: Pair Betting
$1 if and only if horse i finishes ahead of horse j
![Page 92: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/92.jpg)
Example: Pair Betting
$1 if and only if horse i finishes ahead of horse j
A<B B<A A<C C<A B<C C<B ABC 1 0 1 0 1 0 ACB 1 0 1 0 0 1 BAC 0 1 1 0 1 0 BCA 0 1 0 1 1 0 CAB 1 0 0 1 0 1 CBA 0 1 0 1 0 1
![Page 93: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/93.jpg)
What are “reasonable” prices?
![Page 94: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/94.jpg)
What are “reasonable” prices?
For complete markets…
![Page 95: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/95.jpg)
What are “reasonable” prices?
For complete markets…
€
pi =1i∑
![Page 96: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/96.jpg)
What are “reasonable” prices?
For complete markets…
For pair betting…
€
pi =1i∑
![Page 97: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/97.jpg)
What are “reasonable” prices?
For complete markets…
For pair betting…
€
pi =1i∑
€
pi< j + p j< i =1
![Page 98: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/98.jpg)
What are “reasonable” prices?
For complete markets…
For pair betting…
€
pi =1i∑
€
pi< j + p j< i =1
€
1≤ pi< j + p j<k + pk< i ≤ 2
![Page 99: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/99.jpg)
What are “reasonable” prices?
For complete markets…
For pair betting…
€
pi =1i∑
€
pi< j + p j< i =1
what else?
€
1≤ pi< j + p j<k + pk< i ≤ 2
![Page 100: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/100.jpg)
What are “reasonable” prices?
For complete markets…
For pair betting…
In general…
€
pi =1i∑
€
pi< j + p j< i =1
€
1≤ pi< j + p j<k + pk< i ≤ 2
what else?
???
![Page 101: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/101.jpg)
An Axiomatic Approach Path independence: The cost of acquiring a bundle r of
securities must be the same no matter how the trader splits up the purchase.
![Page 102: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/102.jpg)
An Axiomatic Approach Path independence: The cost of acquiring a bundle r of
securities must be the same no matter how the trader splits up the purchase. Formally,
Cost(r + r’ | r1, r2, …, rt) = Cost(r | r1, r2, …, rt) + Cost(r’ | r1, r2, …, rt, r)
![Page 103: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/103.jpg)
An Axiomatic Approach Path independence: The cost of acquiring a bundle r of
securities must be the same no matter how the trader splits up the purchase. Formally,
Cost(r + r’ | r1, r2, …, rt) = Cost(r | r1, r2, …, rt) + Cost(r’ | r1, r2, …, rt, r)
This alone implies the existence of a cost potential function!
Cost(r | r1, r2, …, rt) = C(r1 + r2 + … + rt + r) – C(r1 + r2 + … + rt)
![Page 104: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/104.jpg)
An Axiomatic Approach • Existence of instantaneous prices: C must be continuous
and differentiable
![Page 105: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/105.jpg)
An Axiomatic Approach • Existence of instantaneous prices: C must be continuous
and differentiable
• Information incorporation: The purchase of a bundle r should never cause the price of r to decrease
![Page 106: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/106.jpg)
An Axiomatic Approach • Existence of instantaneous prices: C must be continuous
and differentiable
• Information incorporation: The purchase of a bundle r should never cause the price of r to decrease
• No arbitrage: It is never possible to purchase a bundle r with a guaranteed positive profit regardless of outcome
![Page 107: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/107.jpg)
An Axiomatic Approach • Existence of instantaneous prices: C must be continuous
and differentiable
• Information incorporation: The purchase of a bundle r should never cause the price of r to decrease
• No arbitrage: It is never possible to purchase a bundle r with a guaranteed positive profit regardless of outcome
• Expressiveness: A trader must always be able to set the market prices to reflect his beliefs
![Page 108: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/108.jpg)
An Axiomatic Approach Theorem: Under these five conditions, costs must be
determined by a convex cost function C such that
{∇C(q) : q ∈ RK} ≈ Hull(ρ)
![Page 109: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/109.jpg)
An Axiomatic Approach Theorem: Under these five conditions, costs must be
determined by a convex cost function C such that
{∇C(q) : q ∈ RK} ≈ Hull(ρ)
[
reachable price vectors
![Page 110: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/110.jpg)
An Axiomatic Approach Theorem: Under these five conditions, costs must be
determined by a convex cost function C such that
{∇C(q) : q ∈ RK} ≈ Hull(ρ)
[
10 0 5.5 0 17 0 .9 .9 .9 0 .9 .9 0 42 0 10 10 10 0 0 11.5 8 0 0 1 0 0 0 0 1
securities
outcomes
reachable price vectors
![Page 111: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/111.jpg)
An Axiomatic Approach Theorem: Under these five conditions, costs must be
determined by a convex cost function C such that
{∇C(q) : q ∈ RK} ≈ Hull(ρ)
[
1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1
outcomes
securities
reachable price vectors
![Page 112: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/112.jpg)
Cost Functions Via Duality & The Connection to Online Learning
![Page 113: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/113.jpg)
• Fact: A closed, differentiable function C is convex if and only if it can be written in the form
C(q) = sup x⋅q – R(x)
for a strictly convex function R called the conjugate.
How do we find these cost functions?
x ∈ dom(R)
![Page 114: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/114.jpg)
• Fact: A closed, differentiable function C is convex if and only if it can be written in the form
C(q) = sup x⋅q – R(x)
for a strictly convex function R called the conjugate.
Furthermore, ∇C(q) = arg max x⋅q – R(x)
How do we find these cost functions?
x ∈ dom(R)
x ∈ dom(R)
![Page 115: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/115.jpg)
• Fact: A closed, differentiable function C is convex if and only if it can be written in the form
C(q) = sup x⋅q – R(x)
for a strictly convex function R called the conjugate.
Furthermore, ∇C(q) = arg max x⋅q – R(x)
To generate a convex cost function C, we just have to choose an appropriate conjugate function and domain!
How do we find these cost functions?
x ∈ dom(R)
x ∈ dom(R)
![Page 116: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/116.jpg)
But how do we choose R?
![Page 117: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/117.jpg)
But how do we choose R?
We can borrow ideas from online linear optimization (or the simpler expert advice setting) and in particular, Follow the Regularized Leader algorithms • Market’s conjugate function ≈ regularizer
![Page 118: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/118.jpg)
Learning from Expert Advice Suppose we would like to choose actions based on the
advice of n “experts” (people, algorithms, features…)
![Page 119: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/119.jpg)
Learning from Expert Advice Suppose we would like to choose actions based on the
advice of n “experts” (people, algorithms, features…)
![Page 120: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/120.jpg)
Learning from Expert Advice Suppose we would like to choose actions based on the
advice of n “experts” (people, algorithms, features…)
At each round t, • Algorithm selects weights wt,1 wt,2 wt,3
![Page 121: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/121.jpg)
Learning from Expert Advice Suppose we would like to choose actions based on the
advice of n “experts” (people, algorithms, features…)
At each round t, • Algorithm selects weights
• Experts suffer a loss
wt,1
lt,1
wt,2
lt,2
wt,3
lt,3
![Page 122: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/122.jpg)
Learning from Expert Advice Suppose we would like to choose actions based on the
advice of n “experts” (people, algorithms, features…)
At each round t, • Algorithm selects weights
• Experts suffer a loss
• Algorithm suffers a loss
wt,1
lt,1
wt,2
lt,2
wt,3
lt,3
wt ⋅lt
![Page 123: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/123.jpg)
What is the goal? • Ideally, we’d like to bound the cumulative loss
€
wt ⋅ l tt=1
T
∑
![Page 124: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/124.jpg)
What is the goal? • Ideally, we’d like to bound the cumulative loss • Instead, we look at the algorithm’s regret
€
wt ⋅ l tt=1
T
∑
€
wt ⋅ l t − minw∈Kt=1
T
∑ w ⋅ LT
![Page 125: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/125.jpg)
What is the goal? • Ideally, we’d like to bound the cumulative loss • Instead, we look at the algorithm’s regret
€
wt ⋅ l tt=1
T
∑
€
wt ⋅ l t − minw∈Kt=1
T
∑ w ⋅ LT
algorithm’s loss
![Page 126: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/126.jpg)
What is the goal? • Ideally, we’d like to bound the cumulative loss • Instead, we look at the algorithm’s regret
€
wt ⋅ l tt=1
T
∑
€
wt ⋅ l t − minw∈Kt=1
T
∑ w ⋅ LT
algorithm’s loss
loss of the best fixed weight vector in hindsight
cumulative loss vector
![Page 127: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/127.jpg)
What is the goal? • Ideally, we’d like to bound the cumulative loss • Instead, we look at the algorithm’s regret
€
wt ⋅ l tt=1
T
∑
€
wt ⋅ l t − minw∈Kt=1
T
∑ w ⋅ LT
![Page 128: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/128.jpg)
What is the goal? • Ideally, we’d like to bound the cumulative loss • Instead, we look at the algorithm’s regret
• Can achieve optimal (O(T½)) regret with Follow the Regularized Leader
€
wt ⋅ l tt=1
T
∑
€
wt ⋅ l t − minw∈Kt=1
T
∑ w ⋅ LT
cumulative loss regularizer
wt+1 = argminw∈K
w ⋅Lt + R(w)
![Page 129: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/129.jpg)
Online Linear Opt. Market Making
![Page 130: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/130.jpg)
Online Linear Opt.
• Learner maintains weights wt ∈ K over n items/experts
Market Making
• Market maker maintains prices pt∈Π over n contracts
![Page 131: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/131.jpg)
Online Linear Opt.
• Learner maintains weights wt ∈ K over n items/experts
• Items have loss vector lt, cumulatively Lt+1 = Lt + lt+1
Market Making
• Market maker maintains prices pt∈Π over n contracts
• Contracts are purchased in bundles rt, and qt+1 = qt + rt+1
![Page 132: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/132.jpg)
Online Linear Opt.
• Learner maintains weights wt ∈ K over n items/experts
• Items have loss vector lt, cumulatively Lt+1 = Lt + lt+1
• FTRL selects weights
wt+1 = argminw∈K
w ⋅Lt + R(w)
Market Making
• Market maker maintains prices pt∈Π over n contracts
• Contracts are purchased in bundles rt, and qt+1 = qt + rt+1
• Market maker selects prices
€
pt+1 = argmaxp∈Π
p ⋅ qt − R(p)
![Page 133: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/133.jpg)
Online Linear Opt.
• Learner maintains weights wt ∈ K over n items/experts
• Items have loss vector lt, cumulatively Lt+1 = Lt + lt+1
• FTRL selects weights
• Learner suffers regret
€
wt ⋅ l t − minw∈Kt=1
T
∑ w ⋅ LT
Market Making
• Market maker maintains prices pt∈Π over n contracts
• Contracts are purchased in bundles rt, and qt+1 = qt + rt+1
• Market maker selects prices
• MM has worst-case loss
€
pt+1 = argmaxp∈Π
p ⋅ qt − R(p)
€
maxp∈Π
p ⋅ qT − C(qt ) −C(qt−1)( )t=1
T
∑
wt+1 = argminw∈K
w ⋅Lt + R(w)
![Page 134: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/134.jpg)
Online Linear Opt.
• Learner maintains weights wt ∈ K over n items/experts
• Items have loss vector lt, cumulatively Lt+1 = Lt + lt+1
• FTRL selects weights
• Learner suffers regret
€
wt ⋅ l t − minw∈Kt=1
T
∑ w ⋅ LT
Market Making
• Market maker maintains prices pt∈Π over n contracts
• Contracts are purchased in bundles rt, and qt+1 = qt + rt+1
• Market maker selects prices
• MM has worst-case loss
€
pt+1 = argmaxp∈Π
p ⋅ qt − R(p)
€
maxp∈Π
p ⋅ qT − C(qt ) −C(qt−1)( )t=1
T
∑
wt+1 = argminw∈K
w ⋅Lt + R(w)
![Page 135: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/135.jpg)
An Example • Let K = Π = Δn and
€
R(p) = pi log pii∑
![Page 136: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/136.jpg)
An Example • Let K = Π = Δn and
• Then there is a closed form solution for the prices/weights:
€
R(p) = pi log pii∑
pi = exp(qi ) Σj exp(qj )
wi = exp(Li ) Σj exp(Lj )
![Page 137: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/137.jpg)
An Example • Let K = Π = Δn and
• Then there is a closed form solution for the prices/weights:
€
R(p) = pi log pii∑
pi = exp(qi ) Σj exp(qj )
wi = exp(Li ) Σj exp(Lj )
randomized weighted majority / hedge
![Page 138: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/138.jpg)
An Example • Let K = Π = Δn and
• Then there is a closed form solution for the prices/weights:
€
R(p) = pi log pii∑
pi = exp(qi ) Σj exp(qj )
wi = exp(Li ) Σj exp(Lj )
randomized weighted majority / hedge
logarithmic market scoring rule
![Page 139: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/139.jpg)
More on Choosing R • Interesting market properties can be described in terms of
the conjugate…
![Page 140: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/140.jpg)
More on Choosing R • Interesting market properties can be described in terms of
the conjugate…
• Worst-case market maker loss can be bounded by sup R(x) – inf R(x)
x ∈ Hull(ρ) x ∈ Hull(ρ)
![Page 141: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/141.jpg)
More on Choosing R • Interesting market properties can be described in terms of
the conjugate…
• Worst-case market maker loss can be bounded by sup R(x) – inf R(x)
• Information loss (or the bid-ask spread, or the speed at which prices change) can be bounded too
x ∈ Hull(ρ) x ∈ Hull(ρ)
![Page 142: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/142.jpg)
More on Choosing R • Interesting market properties can be described in terms of
the conjugate…
• Worst-case market maker loss can be bounded by sup R(x) – inf R(x)
• Information loss (or the bid-ask spread, or the speed at which prices change) can be bounded too
Gives us a way to optimize trade-offs in market design!
x ∈ Hull(ρ) x ∈ Hull(ρ)
![Page 143: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/143.jpg)
Example: Permutations • Suppose our state space is all permutations of n items
(e.g., candidates in an election, or horses in a race)
![Page 144: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/144.jpg)
Example: Permutations • Suppose our state space is all permutations of n items
(e.g., candidates in an election, or horses in a race)
• Pair bets: Bets on events of the form “horse i finishes ahead of horse j” for any i, j
• Subset bets: Bets on events of the form “horse i finishes in position j” for any i, j
![Page 145: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/145.jpg)
Example: Permutations • Suppose our state space is all permutations of n items
(e.g., candidates in an election, or horses in a race)
• Pair bets: Bets on events of the form “horse i finishes ahead of horse j” for any i, j
• Subset bets: Bets on events of the form “horse i finishes in position j” for any i, j
• Both known to be #P-hard to price using LMSR [Chen et al., 2008]
• The complex market framework handles both
![Page 146: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/146.jpg)
Example: Permutations
Subset bets (“horse i finishes in position j”)
![Page 147: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/147.jpg)
Example: Permutations
Subset bets (“horse i finishes in position j”) • Hull(ρ) can be described by a small number of
constraints:
€
price(j∑ i in slot j) =1
€
price(i∑ i in slot j) =1
![Page 148: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/148.jpg)
Example: Permutations
Subset bets (“horse i finishes in position j”) • Hull(ρ) can be described by a small number of
constraints:
• Easily handled
€
price(j∑ i in slot j) =1
€
price(i∑ i in slot j) =1
![Page 149: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/149.jpg)
Example: Permutations
Pair bets (“horse i finishes ahead of horse j”)
![Page 150: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/150.jpg)
Example: Permutations
Pair bets (“horse i finishes ahead of horse j”) • Hull(ρ) is a bit uglier…
![Page 151: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/151.jpg)
Example: Permutations
Pair bets (“horse i finishes ahead of horse j”) • Hull(ρ) is a bit uglier… • Solution: Relax the no-arbitrage axiom
![Page 152: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/152.jpg)
Example: Permutations
Pair bets (“horse i finishes ahead of horse j”) • Hull(ρ) is a bit uglier… • Solution: Relax the no-arbitrage axiom • Allows us to to work with a larger, efficiently
specified price space
![Page 153: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/153.jpg)
Example: Permutations
Pair bets (“horse i finishes ahead of horse j”) • Hull(ρ) is a bit uglier… • Solution: Relax the no-arbitrage axiom • Allows us to to work with a larger, efficiently
specified price space • But does it increase worst case loss?
![Page 154: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/154.jpg)
Example: Permutations
Pair bets (“horse i finishes ahead of horse j”) • Hull(ρ) is a bit uglier… • Solution: Relax the no-arbitrage axiom • Allows us to to work with a larger, efficiently
specified price space • But does it increase worst case loss? No!
![Page 155: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/155.jpg)
Extensions and Additional Topics
![Page 156: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/156.jpg)
Relaxing No-Arbitrage • Dudík, Lahaie, and Pennock [2012] pushed on the idea of
relaxing no-arbitrage to provide a general constraint generation technique for constructing efficient markets with “approximately consistent” prices
• Used this to implement the WiseQ market which allowed combinatorial bets the 2012 US presidential and senate elections [Dudík et al., 2013]
![Page 157: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/157.jpg)
Continuous Outcome/Contract Spaces
0 cm 20 cm
$1 $0 $0
![Page 158: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/158.jpg)
Continuous Outcome/Contract Spaces
• Can discretize the outcome space ex ante, but complexity and worst-case loss grow with the number of outcomes
• Most early attempts to avoid ex ante discretization led to negative results [e.g., Gao and Chen, 2010]
• Chen, Ruberry, and Vaughan [2013] extended the duality framework to markets over continuous outcome spaces, generating markets with bounded worst case loss
0 cm 20 cm
$1 $0 $0
![Page 159: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/159.jpg)
Continuous Outcome/Contract Spaces
• Still lots of work to do here quantifying trade-offs between discretization and specially designed markets!! • Worst case loss • Computational complexity • Granularity of predictions
0 cm 20 cm
$1 $0 $0
![Page 160: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/160.jpg)
Making a Profit / Adaptive Liquidity • We have assumed that the market maker is willing to take
a potential (bounded) loss in order to obtain information
• The ideas presented here can be modified to yield market makers guaranteed to earn a profit if the volume of trades is sufficiently high and traders disagree [e.g., Othman & Sandholm, 2011; Li & Vaughan, 2013]
• Yields markets with adaptive liquidity
• In the complete market setting, this requires that prices sum to something more than one – adds some ambiguity when backing out probability estimates
![Page 161: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/161.jpg)
Markets & Variational Inference • The math behind these markets also parallels the math
behind variational inference
mean parameter ↔ prices natural parameter ↔ quantity vector sufficient statistics ↔ payoff function
• This connection can be used to design new scoring rules
[e.g., Lahaie, working paper, 2012]
![Page 162: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/162.jpg)
Price Convergence & Aggregation • When do security prices converge, and do they reflect the
private information or beliefs of the traders?
• Ostrovsky [2012] showed that prices generally converge and incorporate traders’ private information if traders are risk neutral and Bayesian with a common prior
• Price convergence also occurs for risk averse traders with heterogeneous beliefs and budgets [Sethi and Vaughan, working paper 2013]
• Lots to do here too! Which model(s) are reasonable? What happens in real markets?
![Page 163: prediction, belief, and markets - PBworksaaaimarketstutorial.pbworks.com/f/prediction_belief_and_markets.pdf · The Most (in)Famous Prediction Market • 1999: Intrade founded by](https://reader036.fdocuments.net/reader036/viewer/2022070714/5ed58947d42aa526d825d08a/html5/thumbnails/163.jpg)
See the tutorial website for additional references:
http://aaaimarketstutorial.pbworks.com