Private evidence recorder
Create a reproducible field test
Record one changed variable, the controls you held fixed, every full attempt and an honest limitation. Nothing is sent to a server.
Current evidence record
AEFT-20260718-01EVKDE
Draft evidence
Still missing: subject, version / patch, test date, mode, stage / act, fixed conditions, changed variable, observed metric, metric unit, measurement method, at least one full attempt.
Labels describe record completeness and sample size only. They do not certify statistical significance, official accuracy or a universal result.
Optional publication packet
Share evidence without surrendering privacy
This page never uploads your record. These checks create a Markdown packet you may choose to send elsewhere; they do not submit it to us.
Private test library
Saved records and comparability check
No saved field tests yet
Save the current record to build a private test library in this browser. No login or Roblox account connection is used.
Honest confidence language
What each evidence grade means
Draft
A required control, measurement method or one-line record for every declared attempt is missing. Save it, but do not use it as a conclusion.
Exploratory
The context and raw observation count are complete, with 1–4 full attempts. Useful for finding the next question.
Directional
The context and raw observation count are complete, with 5–9 full attempts. A pattern may guide another test.
Repeatable
The same player recorded at least 10 full attempts under the stated conditions. A different player must still reproduce it.
Ten attempts is a practical first repeatability threshold, not a universal sample requirement. Rare drops, high variance, changing enemies and timing-sensitive mechanics can require far more observations. In NIST terminology, repeatability holds the observer and conditions steady; reproducibility requires agreement after specified conditions change.
Before the run
Freeze everything except one variable
- 1. Name the exact unit, team or route.
- 2. Copy the visible version or patch label and test date.
- 3. Record mode, map, Act and difficulty separately.
- 4. Freeze level, evolution, upgrades, equipment, allies, placement and speed settings.
- 5. Define where measurement starts, stops and resets before seeing a result.
After the run
Keep the inconvenient outcomes
- 1. Write one raw observation line for every full attempt.
- 2. Keep zeroes, failures and outliers instead of restarting the sample.
- 3. Use one named metric and unit; do not blend damage, control and economy.
- 4. State timing error, unknown defense or other limitations.
- 5. Start a new record after a material balance or condition change.
Why this exists
A field record fills the gap between a guide and a spreadsheet
Current Anime Expeditions search results include farming advice that tells players to record several runs in a spreadsheet, while other pages publish fixed hourly yields or material-rate labels without a versioned sample trail. This lab turns the useful part of that intent into a structured workflow: conditions, failures, sample size and limitations travel with the result.
Use the Damage Lab when you need transparent build math, the Farming Calculator for failure-adjusted route ranking, and the Raid Planner to diagnose readiness. Export the decisive observation here so its context survives.
Privacy and portability
Your library stays in this browser
Saved records use browser local storage and are limited to the latest 50 on this device. They are not attached to a Roblox identity and are not transmitted to our server. Clearing site data or changing browsers removes the local library, so export important records as Markdown or JSON.
The JSON carries schema version 2 and a deterministic record ID based on normalized test content. Version 1 records migrate locally and remain Draft until their missing method and raw attempt ledger are added. A sharing packet is created only after explicit rights, identifier-review and CC BY 4.0 checks.
Research boundary
Sources and deliberate limits
The verified Roblox experience establishes the game entity, but it does not publish a complete stat export or player-test protocol. A current independent farming guide recommends recording multiple runs, while a separate maps page illustrates the competing fixed-yield style. These establish player intent and the evidence gap; neither supplies official values to this tool.
Protocol language follows NIST measurement terminology: conditions and changed conditions must be stated. The optional sharing gate adapts the general privacy principles in the NIH data-sharing guidance—clear permission, de-identification and limits that travel with shared data. This is a game-observation tool, not human-subject research.
Frequently asked questions
How many Anime Expeditions test runs do I need?+
One to four complete attempts are exploratory, five to nine are directional, and ten or more earn this tool's Repeatable label only when every attempt has a raw observation line. Those are record-quality labels, not statistical proof; rare drops and high-variance mechanics need larger samples.
What makes two field tests comparable?+
The test type, subject, version, mode, stage, fixed conditions, metric and metric unit must match. The changed variable may differ. Notes and limitations can still reveal reasons not to claim a winner.
Does the field test lab upload my data?+
No. Optional saved records remain in this browser's local storage. Copy and download actions happen only when you press them, the publication packet is not submitted to us, and there is no Roblox account connection.
Can a Repeatable record prove a unit is best?+
No. It means the same observer recorded at least ten attempts with the required context and raw ledger. The result remains personal to those conditions and version until other players reproduce it, and another map, team, patch or larger sample can reverse it.