Every resolved verification is scored against the prediction market that settled it. The numbers on this page come straight from the engine's public metrics and update automatically as markets resolve. No backtests, no cherry picking.
Loading the live record…
Correct calls over time
daily result · cumulative record
Brier score over time
NFA (solid) vs market (dashed) · cumulative mean · lower is better
Beat market over time
cumulative share closer than the market · hover for the day
Daily edge vs market
market Brier − NFA Brier per day · above the line = NFA closer
Accuracy by market category
categories with 10 or more resolved verifications · Brier vs the market · best calibrated first
How this is measured
A verification counts as correct when the outcome NFA leaned to is the one that resolved. On multi-outcome markets it counts when the top pick won. The record covers every scored verification except runs the engine itself flagged as no-go before resolution: those carry explicit do-not-trade advice and are tracked separately.
The headline covers the engine as currently configured, from the date shown above. When the engine changes in a way we consider a new version, the record starts a new chapter: the previous chapter's numbers stay disclosed on this page, never mixed into the current one and never deleted.
The Brier score is the average squared gap between the forecast probability and what actually happened: 0 is perfect, 0.25 is a coin flip, lower is better. The market's Brier is computed from its own prices at verification time, over the same markets with the same formula, so the two numbers are directly comparable. "Beat market" is the share of resolved verifications where NFA's probability ended up closer to the outcome than the market's price.
Data-anchored verifications are runs where a structured live data source, such as official statistics, exchange feeds or regulatory filings, served the exact metric the market resolves on.
