A prediction market is a competitive betting game designed to predict specific future events or quantities by tapping into the collective intelligence of a large group of participants. Prediction markets combine many diverging viewpoints, expressed as probability forecasts, into a single probability. This collective estimate changes in real time according to all the available information to human minds.

Individual forecasters place bets on outcomes and receive payouts based on their success. The best forecasters are given more weight in the final mix.

Businesses and governments have used prediction markets in various settings:

• Intel, Hewlett Packard, Ford, Eli Lilly, and EDF (Electricité de France) have used prediction markets to anticipate KPIs, market shifts, sales volumes, a product’s chances of success, and delays in projects.
• The US intelligence community used prediction markets in geopolitical forecasting to understand the likelihood of election results, wars and diplomatic events.
• In the public health sector, the Johns Hopkins Center for Health Security has used crowd forecasting to anticipate the spread and severity of infectious diseases like dengue, malaria and Covid-19.

Prediction markets are most useful when forecasting short to medium term observable events (under 24 months), especially in the following cases:

When the past becomes irrelevant:

Traditional forecasting methods such as time series forecasting rely on the past to predict the future, the assumption being that the past can inform the future. 

But sometimes you’ll have little relevant or reliable data at your disposal to make useful projections, and hanging on to historical data may skew your forecasts. Predictions markets aggregate all the available relevant data using the wisdom of crowds.

When knowledge is decentralized:

Nobody knows everything, but everybody knows something. Usually, we solve hard problems by asking an expert: an engineer, a doctor, a lawyer. But sometimes when a problem is too complex,  when too many variables are involved for a single expert to handle, or when there is too little data to train an artificial intelligence individual expertise falls short.

Prediction markets (or prediction polls like our Prescience platform) help consolidate the informed guesses of the many based on all the available data.

When the situation is fluid:

Forecasts need to integrate new information continuously, you need real time forecasts to be able to react accordingly. 

Human forecasters excel at integrating new information because they spot things that AI would miss, information available on the ground but not yet in data bases.

Prediction markets (or prediction polls) offer 3 main advantages for companies and governments seeking a clearer view of the future:

• They shine under uncertainty. They excel at accurate prediction where other methods fail because they do not rely on AI crunching structured data, but on the informed guesswork of a diversely minded crowd of humans. (hence “crowd forecasting”)
  
  
• They offer precise predictions expressed as probability forecasts. Saying an event is “probable” means different things to different people. “Probable” could mean a 60% chance to one person, but 85% to another. Numbers make things crystal clear, so you can make an informed decision.
  
  
• They are a sturdy and trustworthy forecasting method. By gathering information from many minds thanks to an objective aggregation process, prediction markets easily outperform biased or noisy individual judgments. Relying on a crowd helps reduce the risk of listening to the wrong person, instead trusting a decentralized network of independent points of view.

The first modern prediction market began in 1988 as an academic research project at the University of Iowa’s Tippie College of Business and offered forecasts on that year’s US presidential election. The World Wide Web soon enabled the launch of larger prediction markets targeted at the general public, sometimes bearing other names, like betting exchanges or idea futures.

Hypermind’s pioneering prediction market was first launched in 2000 under the name NewsFutures.

Over the years, this form of “crowd wisdom” has acquired an impressive track record of accurate forecasting in diverse fields ranging from sports and film to business, elections, geopolitics and even medicine.

Prediction Market forecasts are probabilistic. They answer the question: “What are the chances that this or that outcome will come true?” The best way to assess accuracy in the absolute is over many events, by comparing the probability estimations against observed event frequencies.

Here’s a comparison at every level of probability from 1% to 99%, over 5 years, on 400 questions with 1185 possible outcomes, on international topics ranged from elections, geopolitics, geo-economics, and business.

The figure allows us to answer the following question: “Of all the events predicted by Hypermind to be x% likely, what percentage actually occurred?” The closer to x% the answer; i.e., the closer the data points sit to the chart’s diagonal (bottom left to top right), the more accurate the market’s probabilities.

You can use prediction markets to forecast four types of questions:

• binary: “will John Doe win the election ?” (yes/no)
• discrete: “who will win the election ?”(John, Jane, Kane)
• ordered: “what will John Doe’s share of the vote be ?” (more than 55% / 50 to 55% / 45 to 50% …)
• linear: “what will John Doe’s share of the vote be ?” (market price = predicted vote share in %)

They have been used to predict events in various application domains like sports, elections, geopolitics, medicine, science and technology.