Like many mid-sized US cities, El Paso County, Texas (Pop. 850,000) is bracing for the impact of COVID-19, which is still very much on the rise.

As is too often the case, local public health leaders and organizations find it challenging to get access to reliable models and forecasts that can meaningfully support response planning before the crisis hits.

In this question series we ask a set of approximately a dozen core questions that have been formulated in partnership with administrators at Texas Tech University Health Sciences Center, El Paso, and will help guide their on-the-ground decisions.

And there's a contest, to help focus effort and attention on this important series. This time, it's a bit different from the usual. Here, the aim is not just to generate accurate predictions, but to show your work. This is an experiment where we hope to encourage the public sharing of models developed by Metaculites, in order to enable peer feedback and improvement, as well as ultimately a set of tools that could be useful in forecasting similar questions for additional hospitals in need of decision support.

The prize allocation is as follows:

• $1000 for first prize
• $500 for second prize
• $250 for third prize

Prizes will go to users who:

• Publicly post their prediction and updated predictions whenever they make them, and

• Explicitly provide enough information in the prediction, or in an accessible offsite model, for others to understand what exactly went into it, and

• Formulate the model/prediction in a way that it is maximally translatable to other mid-sized (Pop. 250K-1M) US or European cities and their hospitals.

• Submissions will be accepted until 6 May 2020, midnight EST.

Again, the idea here is that if these models are solidly constructed and work well, they can be re-used and re-aggregated for more locales.

Thank you in advance from all of us at Metaculus, and at TTUHSC El Paso, for the modeling efforts you all can come up with!

Fine print:

Judgement will be by the pandemic.metaculus.com moderators, on the basis of the criteria listed above, as well as the model's actual performance against reality. A condition of the prize is that the modeling be either easily reproducible or open-source. There are no formal requirement on the type of modeling done, but it is strongly preferable that the model run on software that does not require a license to run. Metaculus does not promise any technical support, but we will attempt to make some tools available for porting probability distributions out of and into Metaculus.