Does crypto news move prices or follow them? A timestamp audit

A date cluster is a data-quality warning

We counted 3,109 distinct announcement URLs across 11 categories in a saved OKX announcement census. Of these, 978 have an API pTime falling on 3 March 2023 UTC: 31.46% of the census. The internal archive review associates this cluster with help-centre migration and finds earlier dates in article bodies. The count here is a fresh check of stored metadata, not a fresh verification of every original publication date.

If old documents are aligned to a migration timestamp, a price study may attach them to the wrong market period. A date-only label creates another limit: it does not support a minute-precision comparison. An ambiguous timestamp should remain ambiguous rather than be assigned midnight and presented as exact.

Keep four clocks separate

Original publication time records when a publisher says a document appeared. First-seen time records when a particular collector obtained it. Effective time records when a listed action begins. Price observation time records when a market measurement was made. A listing notice, trading start and the site’s first detection can be separated by hours or days.

A collector’s delay is not necessarily a media reporting delay. An edited article may preserve an earlier date, and a republished article may carry a later one. The audit trail should retain the URL, observed metadata, time zone, precision and revision evidence instead of collapsing these clocks into one field.

Why “nearby news” is not a causal finding

A report appearing before a move is compatible with causation, but also with anticipation, another shared catalyst or coincidence. A report appearing after a move may be reacting to prices. Timing is necessary evidence for many causal stories, but timing alone cannot choose among them.

A defensible event study needs a fixed event definition, auditable time alignment, an explicit comparison market and handling for overlapping announcements. It must also distinguish one announcement affecting several assets from several independent announcements. Repeated windows around the same story do not create independent observations.

What we can publish today

The census supports a concrete timestamp-quality finding. It does not support a statement that crypto news generally lags prices by a particular number of minutes, nor a return distribution after hacks, regulation or listings. Such estimates need separately validated event-level samples. On this site, post-detection returns describe observed market changes after collection, with missing windows left unavailable.

Why can a crypto price move before the news appears?

Possible explanations include earlier public information, anticipation, publication delays and collection delays. This archive audits timestamp quality; it does not estimate a typical news delay or identify the cause of every move.

Does nearby crypto news prove what caused a price move?

No. Timing can support or weaken an explanation, but overlapping market information and uncertain publication times remain. The source’s publication time and the site’s first detection must stay distinct.

Limitations

This is an audit of one saved announcement corpus, not a census of all exchange announcements or media coverage. A timestamp cluster is not, by itself, proof of its technical cause. No average reporting lag or causal price impact is estimated.

Sources and data access

• OKX announcement centre

Cited sources document external facts or data formats; they do not endorse our findings. The public snapshot contains the table and methods. Source-file fingerprints are included for recalculated datasets. Full archived raw files are not distributed here; fingerprints identify the files used, rather than independently proving the results.

Aggregate data snapshot (JSON) · Markdown · Methods and limitations

Live prior 30-day hourly baseline — separate from the historical study

Only the current watch list is shown, up to 12 assets. Missing or stale estimates remain unavailable. Venue and quote currency are kept separate; beta needs matching BTC data. Hourly statistics are not minute-move probabilities.

Observation cutoff (UTC): 20/10/2026, 14:59

Related research