How it works
Each question has a precise resolution rule and a closing date. To forecast, add your GitHub handle, a probability and the date to the question’s file. The tool on this page writes the exact lines for you. When a question resolves, every forecaster’s latest forecast is scored.
Scoring
Forecasts are scored with the Brier score: the square of the difference between your probability and what happened, where YES counts as 1 and NO as 0. Lower is better. A perfect forecast scores 0. Always saying 50 percent scores 0.25. Being confidently wrong scores close to 1. The leaderboard averages each forecaster’s scores across resolved questions.
Why forecasting belongs in charitology
Charity has almost no feedback loop: predictions about what will work are rarely written down, so nobody learns whose judgment to trust. Public, scored forecasts are a cheap way to build that track record, a small treatment for the Missing Feedback Loop.
What it isn’t
It isn’t a betting market. No money changes hands, and no personal data is collected beyond the handle you choose to publish.