How to check a market analysis: 4 checks you can do yourself

1. Is there a publication timestamp, before the outcome?

An analysis without a timestamp can always be rewritten afterwards. The first check is simple: issue date and time, in a stated time zone, written on the page. Meridian Engine readings carry the time in UTC: the gold reading was issued on 24 September 2026 at 00:28 UTC.

There is an honest limit: the registration time is the one thing a page cannot prove on its own, and it has to be taken on trust. That is why the conditions are published in full, rather than summarised once the outcome is known.

2. Are the conditions precise enough to be checked?

"The market could fall" cannot be checked. "An hourly close above 4,318.9" can. A checkable condition has a price, a timeframe (the hourly bar, the four-hour bar) and says whether the close counts or the price touched during the bar. Conditions written loosely enough to fit any outcome are worth nothing, even with a timestamp.

In the gold reading there were two invalidation conditions: an hourly close above 4,318.9, or a sweep of the 4,273.8 shelf followed by a close back above 4,304.9.

3. Can the outcome be checked on a public series, separate from the author's?

If the outcome can only be checked on the author's own data, the check is not independent. You need public price bars that anyone can download, and a stated tolerance, because different sources never give identical prices.

The gold reading was checked on two series that are not the feed it was produced from: Binance PAXGUSDT and COMEX gold futures. Closes track the reading's feed within half a point, the extremes of a bar can differ by up to about four points, and every margin reported is wider than that.

4. Do the misses stay published?

Showing only the analyses that worked can make any method look good. The check is whether the archive also holds the invalidated analyses, with the same prominence as the others, and whether aggregate figures carry the number of analyses and the period next to them. A percentage without a sample says nothing.

In the readings record, the BTCUSD reading of 9 August 2026, invalidated at 02:00 UTC by a condition stated at 00:09 UTC, sits beside the gold one.

The four checks on the published readings

The first two are marked "Chosen with the outcome known": they were chosen after the outcome, and stay separate from the readings recorded in advance. The third, the BTCUSD no-trade reading of 1 October 2026, is marked "Recorded before the outcome": it was recorded at 00:25 UTC and published with its outcome, whatever it turned out to be. The public record starts on 10 October 2026.

Frequently asked questions

Does a verified analysis prove the next one will go the same way? No. A reading verified on public data is no proof of the next; only a long record, with every reading in it, counts.

Why does the close count and not the high or low touched? Because the close is a single number per bar, the same on every chart of the same series; the extremes touched during the bar vary more from one source to another.

Which time zone should be used? UTC, the only one that does not change with daylight saving. Meridian Engine readings refer every time to the close of the bar in UTC.

Terms used in this guide

Band · Liquidity pool · Sweep · Displacement · Drift · Shelf

The full method is on the methodology page.