Prediction |

Version Compare

Back to page history

Version User Scope of changes
May 25 2011, 3:44 PM EDT (current) crowdforum 1 widget added, 1 widget deleted
May 30 2008, 2:17 PM EDT crowdforum

Changes

Key:  Additions   Deletions
A complex system such as the outcome of a sporting event or how well a company's sales might perform involves inputs from a wide range of people. A maintenance worker at the event venue might know the condition of the field, a warehouse worker might be aware of conditions that would limit a company's ability to ship product. When all theses points of view are taken into account the crowd can be an excellent predictor of the future.

Bayes Theorem

In layman's terms, Bayes Theorem states that probabilities change as we learn more about a given situation. Bayes Theorem has been used successfully to calculate probabilities of group observations. One of the most famous examples was in the prediction of the final resting place of the USS Scorpion, lost at sea in 1968. Using multiple interviews of people with knowledge of submarine operations, researches were able to piece together a probable outcome for the Scorpion. This same process can be used to predict complex outcomes for business, government & education.

Decision Market

A decision market is a stock market for ideas. Examples include: tradesports.com, Iowa Electronic Markets, Yahoo Buzz Game.
Often set up as a parimutuel where winners share amount bet.

Examples:

  • The king of all decision markets in terms of numbers of participants and money wagered is sports betting. Mirage betting line is 75% accurate, tradesports is 65%.

  • 1996, HP and Charles Plott, an economist from the California Institute of Technology, set out on a joint research project whose aim was to set up a software trading platform for a prediction market at HP.

The participants selected for the research project consisted of 30 product and finance managers from HP. They were each given $50 in a trading account and allowed to sell and purchase contracts on the levels they estimated quarterly sales would reach. For instance, if a manager thought sales would be in the range of $201 million and $210 million, he could buy a contract that would pay at the end of the quarter if his prediction was correct. If he revised his personal estimate throughout the quarter, he would likely try to sell the first contract and buy another based on the new estimate. These markets were open during lunch breaks and after business hours, and lasted for a week. The managers were allowed to keep profits earned from owning paying contracts (correct estimates) at the end of the market. When trading stopped, the contract with the highest price (i.e., under the highest demand from the managers) was deemed the ﾒestimateﾓ of the market. The HP marketing manager estimated simultaneously and independently estimated quarterly sales, as usual.

  • 30 managers from various departments, given $50
  • At quarter end, highest priced contract was deemed the market estimate.
  • Traditional sales estimates had a 13% average error.
  • Decision market estimates had a 6% average error.

  • The United States government attempted to put information markets to use. In 2001, the Defense Advanced Research Projects Agency (DARPA), a research think tank within the Department of Defense, began funding a project called FutureMAP (later under the name Policy Analysis Market, or PAM).