When you see an accuracy number on Narratick, here is exactly what it means, how we calculate it, and — just as importantly — what it cannot tell you. No finance degree required.
Every time the engine spots a narrative it makes a concrete, testable prediction: this stock should move this direction over this window. That is a call. A call is either right or wrong — there is no wiggle room and nothing to argue about later. Do that thousands of times, keep score of the hits and misses, and you get the accuracy figure.
Directional accuracy asks the blunt question — was the engine on the correct side? Brier score asks the harder one — when it said 70%, did the thing happen about 70% of the time? We track both, because they answer different questions, and we publish the engine's Brier score next to the prediction market's on the same resolved contracts.
This is the part most accuracy claims quietly skip. If everything is going up, calling 'up' looks brilliant even with no skill at all. So before we credit a call we subtract what you would have made just by owning the market (the S&P 500) over the same days.
A call that is right in a week and wrong in three months is not the same as one that is right at both. Every call is graded independently at 1-day, 1-week, 1-month and 3-month horizons, and the scorecards are reported per horizon rather than blended into a single flattering number.
The most important rule: we lock in the price the moment the call is made and grade it only against what happened after. Narrative state transitions are timestamped events, so the record is point-in-time — the engine is scored on what it actually said at the time, not on a backfilled view of it.
Flip a coin five times and you might get four heads — that does not make the coin lucky. A few good calls prove nothing, so every accuracy figure comes with a range showing how confident we can really be, and that range shrinks as the number of graded calls grows.
We grade against end-of-day prices from public market data. Intraday swings inside a day are not captured.
The accuracy number is about direction, not a real trading account. It ignores commissions, taxes, and the gap between the price you see and the price you get.
We are scoring individual calls, not running a fund. Position sizing, when to sell, and risk limits are separate decisions.
Unless noted, beating the market means beating SPY over the same window. Some views use a closer sector benchmark.
A strong history is encouraging, never a guarantee. Markets change, and any strategy can hit a cold streak.
Nothing here is a recommendation to buy or sell. Narratick is a research tool — the decisions are always yours.