Do our ranges hold up?

What is being checked

Every 4 hours, for each market we cover, the platform states how far price typically moves: “about 80% of past moves stayed within ±X over the next 12 hours (and 24 hours)”. That statement is saved the moment it is made. When 12 or 24 hours have passed we compare it with the price that actually printed. If the market was closed and there is no price to compare with, that forecast is voided - it is never counted as a success.

This is a range for the size of a move. It says nothing about direction, and it is not a target or a trade signal. If the ranges are good, about 80% of forecasts should stay inside. Every forecast is counted; none is removed or re-run.

Live results

Since 14 October 2026. 7 forecasts issued so far. A row shows a verdict only after 30 checked forecasts.

By market

What a backtest said first

Before showing this to anyone we checked the same band on past data it had never seen: 3.7 years of hourly Gold and Bitcoin, fitted on 2023-2024 and tested on 2025-2026. The 80% band held its stated coverage: Gold 82% (12h) and 81% (24h), Bitcoin 79% and 81%. It is not uniform across conditions - in quiet Gold markets the band covered about 67%, in busy ones about 90% - and two markets over one period is not a guarantee. The live table above is the real test.

We also tested whether our market-structure reading predicts direction. It does not do so reliably, so we do not present it as a signal. The full write-up is in our research notes; the method is on the methodology page.