Forecast With Calibration

A forecast that cannot resolve is a vibe. Write the event, deadline, resolution source, and ambiguity rule before the number. Start from a relevant base rate; the inside view updates that rate, it does not erase it.

One hit or miss is not calibration. Score a comparable set with a proper scoring rule declared before the outcome. Extreme confidence needs extreme evidence; multiplying dependent sub-events manufactures false precision.

Open probabilistic forecasting methods for elicitation, aggregation, scoring, and update cadence.

Use analyze-causal-inference for whether a past change caused an outcome. Use design-product-experiment when assignment can answer the question. Use analyze-critically for confidence in a present explanation.