Modeling the Invisible Workshop Competition Rules

Overview

The Modeling the Invisible Workshop Forecasting Competition is a collaborative scientific forecasting exercise focused on predicting influenza-associated hospitalization dynamics.

Teams will use released hospitalization data to generate short-term forecasts during a simulated influenza season.

The workshop is intended to encourage:

·        scientific modeling

·        forecasting methodology development

·        collaboration

·        reproducible computational research

·        discussion of uncertainty in epidemiological prediction

Team Structure

·        Teams may contain between 1 and 4 members.

·        Participants may belong to only one team.

·        Teams may use any software, programming language, modeling framework, or analysis workflow.

Competition Structure

The competition consists of multiple sequential forecast rounds.

For each round:

·        Organizers release updated hospitalization data.

·        Teams submit forecasts for the next four weeks.

·        Forecasts are scored automatically.

Each round extends the released epidemic trajectory.

Forecast Targets

Teams forecast:

·        weekly influenza-associated hospitalizations per 100,000 population

·        optional estimates of the effective reproduction number (R0 / Rt)

Forecasts are submitted at weekly resolution.

Time System

The competition uses a normalized 40-week influenza season.

·        Week 1 corresponds to the beginning of the season.

·        All forecast targets use integer week numbers.

·        No calendar dates are used in submission files.

Submission Rules

Each team submits one CSV file per round.

Submission files must:

·        follow the official repository format

·        contain the full predicted trajectory through the released data plus four forecast weeks

·        be committed to the repository before the submission deadline

Forecasts are automatically validated and scored through repository workflows.

Permitted Methods

Participants may use any of the following, including combinations thereof:

·        mechanistic epidemic models

·        statistical forecasting models

·        machine learning methods

·        Bayesian approaches

·        agent-based simulations

·        hybrid or ensemble methods

·        manual parameter tuning

·        external public datasets

·        publicly available software libraries

Collaboration within a team is encouraged.

Prohibited Activities

The following behavior is prohibited:

·        attempting to gain unauthorized access to organizer computer systems

·        attempting to gain unauthorized access to other participants’ computer systems

·        interfering with repository infrastructure or workshop systems

·        intentionally disrupting another team’s work

The workshop operates on an honor system emphasizing scientific collaboration and professionalism.

Scoring

Forecasts are evaluated automatically using held-out ground truth data.

Primary evaluation metrics may include:

·        trajectory accuracy

·        RMSE or normalized RMSE

·        forecast stability

·        uncertainty characterization

Lower scores indicate better forecast performance.

Detailed scoring methodology is documented separately in: docs/scoring-rules.md

Reproducibility

Participants are encouraged to:

·        document modeling assumptions

·        preserve reproducible workflows

·        maintain version-controlled code

·        describe uncertainty sources

The repository itself serves as the official archive of forecasts and scoring outputs.

Data Usage

Released workshop datasets are intended solely for educational and research purposes within the workshop.

Participants should not redistribute unpublished organizer-generated datasets outside the workshop without permission.

Workshop Conduct

Participants are expected to:

·        engage respectfully with other teams

·        share ideas constructively

·        acknowledge uncertainty honestly

·        contribute to a collaborative workshop atmosphere

This workshop emphasizes learning and scientific exploration over competition alone.

Organizer Authority

Workshop organizers reserve the right to:

·        clarify rules

·        resolve ambiguities

·        correct data-release issues

·        modify schedules if necessary

·        remove submissions that violate the competition rules

Any major rule modifications will be documented publicly in the repository.

Final Notes

This workshop is intended as a scientific forecasting exercise, not a cybersecurity competition.

Participants are encouraged to focus their efforts on:

·        epidemiological insight

·        forecasting methodology

·        uncertainty quantification

·        reproducible computational science