What this model would have said, month by month

Does the projection work?

The band on the front page is a claim about the future, so it can be scored. We replay it walk-forward: at every past month we rebuild the projection from the data available THEN, and compare it with what the index actually did six months later. The benchmark is deliberately hostile — “assume nothing changes” is hard to beat on a slow index.

Read the sample honestly: these six-month windows overlap almost completely, so 33 checks are nowhere near 33 independent tests, and they all sit inside one war-period regime. Treat the numbers as a historical range this index has stayed inside, not as proven forecasting skill. The forward ledger below is what will eventually settle that question.

Forward ledger — projections with no hindsight in them

The walk-forward test above still replays a rule written in 2026. This ledger cannot: every projection the site publishes is written down here on the day it appears, and scored when its six months are up. It starts empty on purpose.

Retro-run of the verdict rules

Below: what the trigger rules would have said each month since 2022. Read it as illustration, not validation — the thresholds were registered in August 2026 by people who had seen this data. The projection test above is the part that does not suffer from that.