Sample size
1 resolved engine-vs-market pairs. Every figure on this page is computed from that set, and it grows as contracts settle.
Public track record
Every probability the engine publishes is logged and graded when the contract resolves. This page is that ledger: the engine's Brier score against the prediction market's on the same resolved contracts, how often it was on the correct side, and whether its stated confidence matches reality.
Where the numbers stand
What these numbers do and do not say
Direction and calibration, gross of costs
These are measures of whether the engine was right and whether its confidence was honest. They do not subtract commissions, taxes, or the gap between the price you see and the price you get, so they are not the return of a real trading account.
Point-in-time, not backfilled
Every probability is graded against the price and state recorded when the call was made, not against a later revision of it.
Sample size decides how much to read into it
A handful of resolved claims proves nothing. The counts are published next to every figure precisely so the sample can be judged alongside the number.
Past results do not predict the future
Narratick is a research tool, not investment advice. A strong history is encouraging, never a guarantee.