A pint-sized problem for collective intelligence

Mark Roulston, Senior Data Scientist •

Blog agora pub

On Friday 20th March the U.K. Government ordered British pubs to close as part of its “lockdown” to combat the spread of the coronavirus that causes COVID-19. Almost immediately speculation began on when Britons would once again be able to enjoy their alehouses. As with many forecasting problems, different people are making different predictions, either because they have different information or they are interpreting the same information differently.

We used this situation as the basis for an informal prediction market—using our prediction tool, AGORA (beta)—and we invited beta testers of the new product to join the market to predict when the restrictions would be lifted.

AGORA is a platform that synthesises the judgments of different people into a single actionable and accountable probabilistic forecast. It does this using a prediction market mechanism in which the value of an outcome is equal to the probability of it occurring, and the process of price discovery reveals a market consensus on what these probabilities should be.

The participants in this market are not “experts” in the field of pub reopening; perhaps this is a question for which there are no experts. What the participants can do is digest the steady stream of official pronouncements and unattributed rumours emanating from Her Majesty’s Government, and buy and sell contracts corresponding to the months when Britain’s taverns may be allowed to resume their business. Through this activity—from the flow of often vague and sometimes contradictory information—emerges a structured probabilistic prediction which evolves with the flow of information.

Here is a screen capture of the market today: the price of each outcome is the implicit probability of it occurring.

Screenshot 2020-06-04 at 16.43.07.png

As you can see, the current market view is that there is a 31% chance that pubs will reopen in July and a 28% chance that this will happen in August. There is a 30% chance pubs won’t be allowed to reopen until after August and a 10% chance they will have to remain closed for the rest of the year.

Apart from not having been chosen for expertise, those taking part in this market were given no incentive, apart from personal satisfaction, to make accurate predictions and this should be borne in mind when looking at the probabilities. For some predictions, personal satisfaction or bragging rights among peers might be sufficient, but if you really want to motivate participants to collect new data or develop new predictive models, more tangible rewards might be required.

About prediction markets

When designing a prediction market, careful thought should be given to framing it in a way that is well defined and useful. When asking people to make predictions it is important to specify as precisely as possible what you want them to predict: forecasting the “end of lockdown” for example, would have been too imprecise, as lockdown refers to a raft of measures that are unlikely to be lifted simultaneously.

Become an AGORA beta-tester

We are currently running a number of other demonstration markets on AGORA for our beta testers. Some of these markets attempt to differentiate discrete outcomes, while others generate probability distributions for numerical quantities, such as the number of passengers predicted to be using Heathrow airport in August this year.

Alongside our own markets, beta testers are creating their own prediction markets and inviting others to take part.

To find out more about AGORA, take part in a demonstration market or create your own markets, simply SIGN UP FOR AGORA