Electronic MarketplacesOct. 13, 2011Abe Othman1Four Things21. Prediction Market Basics32. Automated Market Making43. Applications to Finance54. Some Project Ideas61. What’s a Prediction Market?7Intrade8Inkling Markets9Predictalot10Events and Definitions11Payouts: Binary and Linear12Think about Traders13The Fallacy of Rational Traders14Zero-Intelligence Modeling15Gode + Sunder ’9316It has no intelligence, does not seek or maximize profits, and does not observe, remember, or learn.17The Simple Static Model18Epsilon-Tweak Model19The same, attractive equilibrium20Traders in Practice21Gates Hillman Prediction Market220 2 4 6 8 10 12 14 16 0 20 40 60 80 100 120 140 160Information Addition Ratio (IAR)Trader rankIAR vs. Rank23Conclusions about Traders242. Automated Market Making251. Not enough volume26Intrade Higgs Boson272. Market too large28(huh?)29Trader 1 Bids onTrader 2 Bids onEVENTSPACE30When does the market clear?31No tradeandNP-hard32w/ Automated MM333. Run lots of markets on obscure events34Robots:Cheaper than people35Case Study: LMSR36Cost function based market maker37Payout Vector38To go from x to y, payC(y) - C(x)39C(x) = b log sum exp(xᵢ /b)40Red Sox - Yankees41(5,3) → (6,3)b=1042C(y)-C(x) = 0.56243Bounded Loss44b log n45Contrast withfixed prices46Marginal Prices47Prices change instantaneously as bets are made48pᵢ = exp(xᵢ/b) / sum exp(xⱼ/b)49Form a probability distribution!50No-regret online learning51Weighted Majority52My research53How to add in profit, depth, ...54Latest result hasfour properties551. Bounded loss562. Can make a profit573. Zero bid/ask spread (in the limit)584. Unbounded depth (in the limit)59These are hard conditions to mutually satisfy603. Applications to Finance61OperationalizingRisk Measures62Finance was here firstAI is here better63Sharpe, VAR, Others64Example65Sharpeμ/σ66Risk measure view67Given a set of candidate strategies68...the best is the one with highest Sharpe69Automated trader view70Given the chance to make a trade71...do it if it increases your Sharpe72What does this look like in practice?7374Step back75Risk measures are trading agents76Famous risk measures suck77Famous risk measures suck as trading agents78Make bettertrading agents79Make betterrisk measures804. Final Project Ideas811. Improve Predictalot822. Dig through my GHPM data833. Simulate different market makers vs. the same
View Full Document