EVERNESS ATLASUNOFFICIAL NTE GUIDES & TOOLSOpen My AtlasAtlas
Illustrated testing guide

How to Compare NTE Builds

Separate visible stat changes from what actually happens in repeat runs.

Status: Editorial testing method · Updated September 13, 2026
Source status: Atlas tool behavior · Confidence: High for the workflow; no universal build claim
One controlled loop

Move from one build question to a result you can inspect

Use the same order each time. The diagram is a testing sequence, not a score or recommendation.

01

Ask

Name one outcome and one loadout choice you want to examine.

02

Record

Write down both loadouts and every condition that must stay fixed.

03

Read

Compare only the panel values visible in your current client.

04

Run

Alternate A and B, recording three comparable successful runs for each.

05

Review

Read exact values, median, range and your notes before deciding what to retest.

Original Everness Atlas testing diagram. Every value and observation comes from the player.
Before entering numbers

Define one change and freeze the rest

QuestionWhat single choice are you testing?

Name the Arc, module or other loadout field exactly as you see it. Avoid changing several fields and then assigning the result to one of them.

Keep fixedWhat makes the runs comparable?

Record the character, game version, team, encounter, difficulty, skill levels, starting resources and timing points you intend to hold still.

Save both setups in Build Notebook. Private notes remain there unless you deliberately copy a supported field into another tool.

Two different questions

Read panel changes and run outcomes separately

A visible number and an observed clear time describe different parts of the test. Keep both, and keep their limits visible.

RecordWhat it showsWhat it cannot establish
Loadout record

What you equipped and which conditions you intended to keep fixed

Whether the setup performed better

Panel snapshot

Exact visible A/B values, arithmetic difference and percentage change

Damage, uptime, caps, interactions or encounter fit

Repeated runs

Your recorded clear times, median, range and run-to-run variation

A universal ranking or statistical proof from six runs

Run notes

Failures, interruptions, execution differences and reasons to retest

A result unless the observation is recorded consistently

Evidence-reading matrix by Everness Atlas. Limits apply even when one result looks clearly larger or faster.
Panel evidenceBuild Stat Snapshot

Enter exact visible values. Each row compares A and B on its own scale.

Observed outcomeBuild Comparison

Record three successful times per setup, then review median, range and exact runs.

After six recorded runs

Choose the next action, not a universal winner

Clear differenceRepeat once more

Re-run the same conditions before treating the pattern as useful for your account.

Overlapping rangesCollect more observations

Variation may be larger than the difference you are trying to read.

Interrupted or failed runsKeep the notes

Do not silently replace inconvenient attempts. Record why they were not comparable.

Several variables changedReset the test

Create a narrower A/B pair so the next result answers one question.

Interpretation boundary: this workflow produces descriptive evidence for your own account and test conditions. It does not calculate DPS, prove statistical significance or rank a build for every player and encounter.

Build comparison checklist

  1. Write one test question.
  2. Save Build A and Build B with the same version and context.
  3. Copy only visible current-client values into the panel snapshot.
  4. Alternate A and B under the same recorded conditions.
  5. Keep failures and interruptions in the notes.
  6. Review exact values, median and range before planning the next test.

Record two loadouts › · Open the Shinku evidence page ›

Changelog

September 13, 2026 — Published the illustrated five-step build comparison method, evidence-reading matrix and decision paths.