Trust · method

How we test, and what we will not publish.

This site exists to publish things the official pages do not: how long a match runs, how many players are in it, what it pays, and what the client does when something goes wrong. That only works if you can tell our measurements from the developers' marketing.

Three markings

Every mode page and every paragraph of claim on this site carries one of these.

Tested Tested. We saw it happen, in a match, on the date printed next to it, with a screenshot that has a timestamp. If we only saw it once, we say once.

Official only Official only. It is the developer's own wording, repeated as their claim. We have not independently confirmed it and we do not dress it up as a finding.

Needs check Needs check. Something exists — a queue, a counter, a mechanic — but we have not measured the details. These are the honest holes in the site.

What counts as evidence

• The observation and the screenshot must come from the same session.
• Numbers are read off the screen, not reconstructed from memory. If the panel says 3.0 minutes, we publish 3.0 minutes.
• One sample is labelled one sample. Averages wait for more matches.
• Anything derived from several observations is labelled a hypothesis and shown with the inputs, the way the XP rate is on the comparison page.
• Screenshots are kept at full resolution in the repository so a claim can be re-checked later.

What we will not publish

• Prices, drop rates or damage numbers we have not seen in the client.
• Player counts without the timestamp they were read at.
• Anything carried over from a different block game. Miniblox shares a genre with Minecraft, not a rulebook.
• Claims about loot tiers or variants from a single match.

Corrections

If a figure here is wrong, the useful message names the mode, the panel and what it says now. Direct corrections are listed on Sources with the date they were made.

Method published 30 October 2026. First sessions: Survival, EggWars and Skywars on the same day, client v3.46.231.