Timestamped first
Every prediction stores its original wording, probability, target, source, and deadline before it can be scored.
Methodology
Scoring built for accountability
A clear prediction must be specific, time-bound, and recorded before its outcome.
History is preserved
Forecast edits create another audit version. They do not erase the original record or rewrite the publication time.
W/L stays understandable
A call earns a win when its defined outcome meets or exceeds the recorded target. Otherwise it receives a loss.
Confidence has consequences
Brier score is calculated as (probability − outcome)². Lower is better. Forecast Score converts it to an intuitive higher-is-better percentage.
Risk is estimated before resolution
Member analysis simulates target hits, deadline prices, likely ranges, and drawdowns using recent market volatility and the time remaining.
90% call wins99 Forecast Score
60% call wins84 Forecast Score
60% call loses64 Forecast Score
90% call loses19 Forecast Score
Forecast risk and difficulty
The member risk snapshot runs 3,000 neutral-return price paths calibrated to recent daily returns. It uses Geometric Brownian Motion by default and switches to Jump Diffusion only when at least two unusually large moves appear in up to six months of history. The panel reports target-hit probability, expected deadline price, a 5th–95th percentile range, finishing probabilities, expected maximum drawdown, and a difficulty score equal to 100 minus the estimated hit probability.
Why show both?
Win rate is easy to understand, but it treats cautious and highly confident calls alike. Brier scoring rewards forecasters who are both correct and well calibrated, while strongly penalizing certainty that turns out to be wrong.
Legacy predictions
Older imported calls remain visible. If a legacy record lacks a genuine probability or deadline, it can appear in the ledger without receiving an invented confidence-aware score.