What is PredictWise?
PredictWise is a website for a project that studies the collection of individual-level data for predictions, aggregation of that data into prediction, and usage of predictions. PredictWise focuses on public domains, such as: politics, sports, finance, and entertainment. What you see is the continuously running predictions. What you do not see is the research aspect of the work that is housed @ ResearchDMR.com.
What is the point of PredictWise?
David Rothschild started this website, along with Andrew Leonard, when he was a graduate student. There were two goals: (1) Displaying the fruit of his research in real-time ensured more comprehensive data collection and no look-ahead bias in the his methodology. (2) We thought it would be interesting and informative for people to better understand the likelihood of certain major events occurring, and to have that information presented in a manner that is easy to comprehend.
Should I make any bets based on this information?
No, betting money on sports, politics, or anything else is a bad idea. Not only is it losing proposition due to the house's commission, but it is addictive, destructive, and - depending on your location - most likely illegal.
What is the point of PredictWise's data collection?
While the main focus of the site is aggregating data that is already available: markets, polling, etc., sometime we run experimental data collection from the site. It is for academic research.
Who runs PredictWise?
David Rothschild runs PredictWise on a daily basis. If you are interested in blogging or helping in any other way, please email him @ David@ResearchDMR.com.
Who is David Rothschild?
David Rothschild is an economist at Microsoft Research. He started PredictWise, along with his friend Andrew Leonard, as a graduate student at University of Pennsylvania's Wharton School. The site is designed to showcase the fun, public aspects of his academic research. You can learn more about his research @ ResearchDMR.com.
How to I contact David Rothschild?
By email: David@ResearchDMR.com
Overview of Data and Methods
Most of the focus is on prediction market data, polling data, and bookie data. Prediction market data comes from exchanges that host markets where people buy and sell contracts on upcoming events. Polling data comes from surveys designed to reflect the sentiment or expectation of a target population. Bookie data is the odds/price that bookies are willing to sell contracts on upcoming events.
Generically, most tables focus on market data. There is a three step process. Step 1: construct prices from the back/sell, lay/bid, and last transaction odd/price in the order book. We always take the average of the highest price traders are willing to buy a marginal share and the lowest price people are willing to sell a marginal share, unless the differential is too large or does not exist. Step 2: correct for historical bias and increased uncertainty in constructed prices near $0 or $1. We raise all of the constructed prices to a pre-set value depending on the domain. Step 3: normalized to equal 100% for any mutually exclusive set of outcomes.
What is Betfair?
Betfair, based in the United Kingdom, is the world's largest internet betting exchange. Rather than having a bookmaker create odds, the odds for every bet are determined by the market of bettors, working similarly to a stock market. Bettors can either "Back" (buy) or "Lay" (sell) a given bet at certain odds, and the odds move as Backs and Lays are matched. Betfair is legal in the UK and other countries, but it is illegal to bet money on Betfair as a resident of the United States.
What are the Iowa Electronic Markets (IEM)?
The Iowa Electronic Markets (IEM) are a group of real-money prediction markets operated by the University of Iowa Tippie College of Business. The IEM is not-for-profit; the markets are run for educational and research purposes. Because of the small sums wagered and the academic focus, the IEM has received no-action relief from the United States government, meaning U.S.-based speculators can legally risk between $5 and $500 on a wager.
What is PredictIt?
PredicIt is an exchange that operates legal prediction markets for United States citizens. PredictIt has received a no-action relief from the United States government, meaning U.S.-based speculators can legally risk up to $850 in a market.
What is Hypermind?
Hypermind is a play money-market, but has real prizes that correspond to the play-money. Rather than focus on large quantities of users, it tries to use highly trained collections of traders.
What is the Hollywood Stock Exchange (HSX)?
The Hollywood Stock Exchange (HSX) is a play-money prediction market in which users can buy or sell shares in movies, actors, directors, and other Hollywood-related topics. For example, users can buy or sell shares of an upcoming film as a means predicting how well that film will do at the box office in its first four weekends of wide release, and then be ranked based on the accuracy of their predictions. Because HSX involves only simulated money, it is legal for all participants.
What is HuffPost Pollster?
HuffPost Pollster is a site that discusses and aggregates polling data. Polling data is subject to random fluctuations and Pollster's aggregation methods cleanly and transparently aggregate polls over time to provide a more meaningful snapshot of where the polls are at any given moment.
What is Oddschecker?
Oddschecker is a site that posts the latest bookie odds/prices to sell contracts for upcoming events. We generally focus on the highest lowest odds/prices that any of the bookies will sell at, as the marginal sell price from bookies.
Sometimes PredictWise only lists one source but has a slightly different prediction than that source - why is that?
The answer is de-biasing and normalizing. First, we are adjusting for known biases between raw data and outcomes. Second, when the list of possible outcomes is finite (such as the 32 teams that can win the Super Bowl), the sum of every possible outcome's probability should equal 100%. Because our sources must create a separate market for each outcome, this is not always the case. To adjust for this, we divide each outcome's probability by the sum of every outcome's probability and use that as our prediction. We always try to show the raw data for sophisticated users who do not just want to see the polished prediction.
You list a single price from prediction markets like Betfair, but in actuality prediction markets have simultaneous (and different) Bid price, Ask price, and Last price. How do you calculate this?
As long as the spread between the Bid and Ask prices is less than 15 points, we take the average of the two. If not the spread is larger, we may go to the last price. Again, we always try to show the raw data for sophisticated users who do not just want to see the polished prices and predictions.
What happens when a market lacks liquidity and there is a bid, but no ask or an ask, but no bid?
If there is a bid/back that is greater than 50 and no ask/lay, we take the average of the bid/back and 100. If there is an ask/lay that is less than 50 and no bid/back, we take the average of the ask/lay and 0. If neither of those conditions hold, we take the last price. We understand that sometimes this does not yield the obvious answer, but we need the numbers to speak for themselves, so we never adjust any numbers manually. Again, we always try to show the raw data for sophisticated users who do not just want to see the polished prices and predictions.