Pastcasting is the practice of forecasting on a question that has already happened, from a vantage point further in the past.

Unlike forecasting, pastcasting allows for quick feedback even on questions that were open for a long time, which might be systematically different than questions with shorter time horizons. Unlike calibration training, it also tests relevant forecasting skills (trend extrapolation, investigating different views, and determining trustworthiness of news sources).

Vantage Search is a custom search engine that only shows results from before the vantage point to prevent information leaks from the future.

Preliminary results should be available within a few seconds. If you're experiencing a much longer delay, please check your adblocker and/or VPN. Archived results take much longer to load due to rate limiting by the Wayback Machine API.

We use already resolved Metaculus and Good Judgement Open questions as sources. We provide a link to the original question page after submitting a pastcast.

By checking that you have prior knowledge of the question, you can help us improve the quality of our questions for future users.

Pastcasts are scored using log scoring relative to the original crowd forecast at that time, meaning that you will recieve zero points if you submit the same value as the crowd. The scoring rule is also strictly proper, which means that your expected score is maximized if you report your true beliefs.

We're currently gauging whether pastcasting is a useful addition to the forecasting ecosystem (mostly by usage). If it is, we'll commit to spending more time improving the website and adding more features.

You can email us at hello@sage-future.org