How the score works

Interest-rate differential. Each currency's central-bank policy rate is compared with the average rate of the other seven currencies. A currency whose rate is above that average gets support, and one below it is weighed down. We update the rate table after each central-bank decision.

COT positioning. Every week the US Commodity Futures Trading Commission (CFTC) publishes how large traders are positioned in currency futures. We rank each currency's current positioning against the past 52 weeks. We read it as a contrarian signal: when positioning is crowded on one side, the score leans the other way.

Central-bank tone. We read central-bank statements, and news reporting on them, over the last 30 days. Each one is marked hawkish (leaning toward tighter policy) or dovish (leaning toward looser policy). Recent items count more than older ones. Statements without a clear lean do not move this part.

News. An AI model reads incoming financial news and marks each story as positive or negative for the currencies involved, and how much it matters. A story's effect fades over time. Repeats of the same story are cut down, so one event is not counted several times over.

Market backdrop. We also check whether markets are broadly risk-on or risk-off, and adjust how far the score sits from neutral to match.

Between updates the score carries over, and without new information it drifts slowly back toward 50.

The whole score is recalculated every hour. Its inputs change at different speeds:

• News is collected every hour.
• Central-bank tone changes when a new statement or report appears.
• COT positioning changes once a week, when the CFTC publishes its report.
• Interest rates change only when a central bank changes its rate.

That is why some scores barely move for days and others jump after a big news event.

The score summarises public information. It does not predict prices, and a strong score does not mean a currency will rise. Use it as one input among many.