Skip to content
Doefin

Educational guide

Why difficulty forecasts need a range

A forecast estimates the next difficulty adjustment. Its range describes uncertainty. Assessing accuracy needs recorded forecasts and outcomes.

Guide reviewed

Why a forecast changes

A point forecast is a single estimate of a future difficulty adjustment, based on information available when it was issued. As the remaining blocks arrive, the observed pace can change and so can the forecast. Comparing forecasts therefore requires their issue times as well as their values.

A range needs a method

A forecast range needs an explanation of how it was calculated, which assumptions it uses and how often it is intended to contain the outcome. A range derived from a theoretical model and one fitted to observed data both need evaluation. The label alone does not show that either works.

Accuracy and coverage measure different things

Point error measures the distance between a forecast and an outcome. Interval coverage measures how often outcomes fell inside a published band across a set of forecasts being evaluated. Neither measure gives the full picture on its own. The band’s width, how far ahead each forecast was issued, missing observations and changes to the method also matter.

What evidence of accuracy requires

An accuracy assessment needs the forecasts as originally issued, the source outcomes and a defined evaluation period so that someone else can repeat the comparison. The sample data on this website demonstrates calculations only. It does not establish a calibrated interval, an accuracy ranking or a general claim about competing forecasters.