docs-website/source/uk-challenge/index.md

In epiforecasts/crowdforecastr: Eliciting crowd forecasts in R shiny

title: The UK Crowd Forecasting Challenge subtitle: Win a prize for your insights! layout: page

Overview and summary

From May 24th 2021 until August 16th 2021 we will collect and score forecasts of Covid-19 case and death numbers of Covid-19 in the United Kingdom. The best forecasters will receive a prize of

1. prize: 100 GBP
2. prize: 50 GBP
3. prize: 25 GBP

All forecasts will be made using the crowdforecastr platform (app.crowdforecastr.org or rt-app.crowdforecastr.org)

Study period

• 12 weeks
• first forecast date: 2021-05-24
• last forecast date: 2021-08-16

Forecast targets

Forecasters are asked to predict - the number of Covid-19 case numbers and - death numbers - in the United Kingdom - on a 1 to 4 week ahead horizon

Eligibility

Everyone who submits at least one forecast for both targets (cases and deaths in the UK) is eligible to win a prize.

In order to receive a prize, you must make your contact details known to the organisers (either by providing your email address in the app user account, or by sending a message to [e-mail])

Making a forecast

Forecasts will be made through the crowdforecastr platform. Individuals can make a forecast using either the regular forecasting app (app.crowdforecastr.org) or the Rt forecasting app (rt-app.crowdforecastr.org) Forecasters can make as many forecasts as they like - only the latest one will be counted

The regular forecast app

• app.crowdforecastr.org
• Forecasters are asked to give the median and width of a chosen predictive distribution for each target

The Rt forecast app

• rt-app.crowdforecastr.org
• Forecasters are asked to give the median and width of a chosen predictive distribution for Rt values
• These Rt values then get mapped to cases as well as deaths using a renewal equation as implemented in the R package EpiNow2
• Users can visualise case forecasts, but will not be able to see corresponding death forecasts and have to trust the renewal equation to produce sensible results.

Evaluation

A ranking will be based on relative skill scores that are obtained using pairwise comparisons between forecasters: - Every forecast will be scored using the weighted interval score (WIS). Scores will be computed based on the log of the forecasts and truth data. This preserves the incentive of the forecasters to give their best forecast, while at the same time giving a more equal weight to death and case forecasts. - for every possible pair of forecasters, a mean score ratio will be computed. The mean score ratio is the mean WIS achieved by forecaster 1 over the mean WIS achieved by forecaster 2 based on the overlapping set of prediction targets for which both forecasters have made a forecast - For every forecaster, a relative skill score will be computed as the geometric mean of all mean score ratios involving that forecaster

If you miss a forecast date you will be assigned the median score that was achieved on that date by all forecasters who have made a forecast

epiforecasts/crowdforecastr documentation built on June 23, 2021, 10:30 p.m.