We've released Ergo, a Python library for forecasting that people can use, provide feedback on, and contribute to.

With Ergo you can

• Import questions and community forecast distributions from Metaculus and Foretold
• Write models that relate different questions to each other by using community distributions as random variables
• Submit model predictions back to Metaculus, automatically converting them into the right format
• Specify distributions using 90% confidence intervals
• Sample from other primitive distributions

To get you started, we've implemented Ergo in a few Colab notebooks with models that predict

• Confirmed case counts
• Lockdown duration
• “True” cases (includes cases that haven't been tested)

Next up on our roadmap are features that let you

• Define and store links across Metaculus questions to sense check predictions or follow certain questions in Colab
• Track community distribution changes

We hope this makes it easier for Metaculites to build models and update them regularly. You can find Ergo documentation here.

Feel free to email us at ergo@ought.org to

• Request features
• Ask questions or provide feedback
• Ask for help building a model if you're not familiar with Python. We’ll consider building a model for you if it’s consistent with the Ergo roadmap

Happy forecasting!