From a question to an upgrade decision

Your first run

1. Choose a question. Select Overload, Skills, Cubes or Dupes. Switching questions keeps your setup. Cubes use Benchmark; your separate Account build remains available when you switch back.
2. Choose the tested unit. “Analyze unit” chooses a physical slot. Use the visible squad cards to change units, cubes and Burst participation. Expand “Target settings” to adjust the battle.
3. Review the baseline. Benchmark starts from standardized Lv400 stats, OL5 gear primaries and compatible SR15 dolls. OL starts at zero effects, Skills defaults to 1/1/1 on the tested unit, Cubes compares against no cube, and Dupes starts at the chosen copies and bond. These are different questions, so their percentages are not interchangeable.
4. Run deliberately. Start with Balanced compute. Normal runs focus on damage comparisons. Research is optional and explicitly starts farming calculations or exhaustive skill searches. Cancel is safe; completed verified replays are kept for resuming the same setup, while incomplete analyses are not recommendations.
5. Read the answer, then inspect it. Look at personal and team improvement, the comparison baseline, and the completed-run assumptions. Use the relevant detailed view before spending resources.

The example you start with

Example version 2026-09-11.1 uses Liter, Crown, Cinderella, Scarlet: Black Shadow and Naga. Cinderella is the tested unit; Naga’s Burst is disabled. It is a control demonstration, not a best-team claim.

The starting target has DEF 100, a core with a 50% target fallback core-hit rate (weapon and slot settings take priority), no parts and effective range enabled. Duration is 180 seconds, with natural Burst cycling and 24 response-delay ticks. All units start without OL effects or cubes. Every comparison is calculated when you run it; no example damage numbers are preloaded.

Changing the tested unit in Advantage mode can change the target element. Use Neutral and keep the same target settings when you need a neutral comparison across different elements.

Overload: which effects are worth farming?

Worked workflow: keep Cinderella selected, choose Overload, review the target and run Quick. Read the best tested effect mix, then open Roll Depth. Compare one, two and three desired effects per piece: that means 4, 8 and 12 effects across the set.

What to do afterwards: compare the damage gained at each line depth and choose how much output you want to retain. Open Research analysis and click Calculate only if you also need farming costs or value-tier returns. Use Custom Compare to replay a specific alternative under that same completed run. Export its OL target into the Custom Module Planner to plan from your actual gear.

Skills: where should the next manuals go?

Worked workflow: switch to Skills with the same team. Keep the tested unit’s starting levels at 1/1/1 to study a fresh investment, or enter your actual levels. Run skill analysis and compare S1, S2 and Burst curves alongside practical mixed builds. A fresh 1/1/1 start normally evaluates 55 builds, including mixed 4/7/10 milestones.

What to do afterwards: find an upgrade with a useful damage increase, then review its manual costs. The normal run includes isolated curves and mixed builds, but it cannot establish a global optimum. Research checks every upgrade combination from your starting levels and adds a connected upgrade sequence: 1,000 combinations from 1/1/1, or 64 from 7/7/7. Already completed replays are reused.

The “best value” endpoint balances modeled gain with normalized material burden; it is not automatically best for your inventory. Open the inventory and income settings inside Research to enter owned manuals and realistic income before interpreting days. Changing these settings takes effect when you calculate Research again; the completed result retains its original assumptions. Skill, Burst and Code manuals remain separate currencies. Levels 4, 7 and 10 are resource landmarks, not guaranteed damage breakpoints.

Cubes: which choice and level help this team?

Worked workflow: switch to Cubes, keep the shared comparison level at 7, and include both Advantage and Neutral. Start with no OL effects. Run the comparison, then read Ranking and Sensitivity.

What to do afterwards: if the leading cube changes between conditions, use the row matching your intended fight. If cubes tie, damage gives no reason to prefer one of those co-winners. Check Level Breakpoints before spending batteries; a level increase can add common stats, a useful effect breakpoint, or very little modeled gain.

No cube is the baseline, including its missing common primaries. Teammate cubes are cleared by default; select Preserve to evaluate your configured teammate cubes. The chart measures outgoing damage. Defensive, healing, accuracy and spatial effects have explicit limitations and may matter outside that number.

Dupes: what does another copy add?

Worked workflow: switch to Dupes and choose the unit you want to discuss. Set its current extra copies and bond. Choose “Dupes only · keep bond fixed,” then run. Read the next-copy personal gain and team gain separately.

What to do afterwards: use the next-copy result for the marginal damage value at that investment. The full curve shows diminishing or changing returns at later tiers. A support unit can add team damage with a modest personal increase; that is why both numbers matter.

“Include unlocked bond” trains the selected unit to its cap at every tier, including the baseline, so that comparison includes additional bond training. It is not immediate damage from acquiring the copy. Only the selected physical slot changes—even if that same character appears elsewhere. Character-specific caps apply. Damage value does not include general pull value, collection goals, level-cap access or combat-power penalties.

Before applying or sharing a result

• Confirm the selected unit, teammates, Burst participation, element, DEF, core rate, range and duration. Support effects can change rankings when the team changes.
• Compare like with like. Benchmark excludes account progression; imported builds can include bond, dupes, recycling and different investments.
• “Exact replay” means the simulator evaluated that candidate. It does not mean the modeled mechanics have all been verified in game. Read provisional-model notes and Beta limitations.
• Changing relevant inputs clears affected results. Run again before using them. Saved results retain the original completed setup.
• Use the on-screen charts and exact tables to inspect your results. CSV exports retain the detailed values, and an OL target can be opened in Custom Module Planner.

If something looks wrong

For unexpectedly low damage, check enabled units, Burst order, target element, core assumptions and whether the character is supported. If the page is slow, keep Balanced compute and try a shorter test before a full search. Lower worker counts change concurrency, not the calculation rules. If storage is blocked, analysis still works for the session; export anything you need to keep.

Account build accepts a simulator scenario JSON; this edition does not add direct cloud Library synchronization. With no import, that workspace initially contains the labeled example. Saved data stays in this browser profile and is separate from the private creator edition. Clearing browser data removes local saves. Page and asset requests still reach the website host. Privacy details.