Keep the source.
Every entry links back to a dated public statement. Summaries are labeled. The collection grows through source submissions and editor searches.
Words now. Evidence later.
The future,
on the record.
People make big predictions.
Let’s remember what they meant.
Read the source. Agree on a clear statement.
Put a probability on it. See how it turns out.
THE FORECAST FILEVOL. 01 / 5 VOICES
A map of their convictions
How sure
did they sound?
Each mark is a forecast.
Its position is the starting interpretation of their confidence.
Unresolved Resolved
Select a mark to read the source ↗
Opening the record…
Connecting to community ratings…
Everyone, on the record.
Source first. Interpretation second.
The accountability part
A forecast is only as good
as the record behind it.
Compare resolved claims, inspect the assumptions, and keep the sample size in view. A small, selected archive is a starting point for evaluation.
Historical illustrationThese starting probabilities were reconstructed after some outcomes were known. They are not an independently established performance ranking. New ratings do not rewrite old scores.
| Public figure | In the record | Resolved | Scorable | Brier loss ⓘ | Prospective score |
|---|
Confidence has consequences.
A 90% forecast that happens has a Brier loss of 0.01. If it fails, the loss is 0.81. Lower is better. Only binary outcomes with usable criteria, deadlines, and probabilities are included.
A prospective record needs the wording and probability frozen before the outcome is known. Conditional claims without agreed resolution criteria stay outside the score.
The Spilled Ink approach
Less “I told you so.”
More “let’s check.”
Agree on what was meant.
Vote on the proposed wording. Then estimate the speaker’s implied probability: lower, about right, or higher. This asks what they meant, not whether you agree.
Let reality settle it.
Record the deadline, resolution rule, outcome, and evidence. Historical scores use the published starting interpretation. A prospective score requires a pre-outcome freeze.
Collection, voting, and scoring details
This is a selected starter archive, not every forecast these people have made. Missing forecasts and source corrections can be submitted for review. Quoted probabilities belong to the speaker; editorial starting probabilities are our interpretations. Claims without a usable time limit or resolution rule remain unscored.
Community confidence is the median numeric interpretation. Wording votes and confidence votes are separate. One random browser token updates one response per forecast; this is not proof of one person, one vote. Clearing browser storage loses access to that response. In preview mode, votes stay only in this browser and are excluded from community totals.
Brier loss is the mean of (probability − outcome)², using probabilities between 0 and 1. Scores cover the scorable subset and do not adjust for forecast difficulty, dependence, topic selection, or hindsight. Resolved retrospective claims do not establish a calibrated public track record.