How this tool works
What it does
The build tool finds character builds under two modes: Soldier (max damage, min cost) and Profiteer (max cases looted per day, net of the $ that loot is actually worth against cost). Both fold combat income into their cost ladder, and both optimize the 9 combat skills plus gear, ammo, food and prestige, on one shared skill-point budget with the Companies and Management levels you set, given your level, rank/battle bonuses, and any pins you set. Profiteer forces the Cases cost factor on for this reason — without it, the optimizer would chase raw case count without any sense of what those cases are worth.
Soldier and Profiteer give you the best build at each net-cost tier rather than a single "best" build, sweeping from −0.10 (a build that turns a $0.10 profit per 1K damage, or per case for Profiteer) up to the maximum-damage (or maximum-cases) build in steps of 0.02.
Every economy skill competes with the combat skills for the same skill points. Management (+2% on your NPC employees, and better training cards), Companies (how many companies stay active), Energy and Production (self-work: your energy bar turned into production points) are searched alongside the nine combat skills, so a build only buys economy levels when the income they bring in beats what the same points would do for damage or cases. Pin any of them to fix its level, or turn a source off under Cost Factors to take its skills out of the search.
The optimization algorithm
The optimizer is NSGA-II (Non-dominated Sorting Genetic Algorithm II), a well-established multi-objective evolutionary algorithm. It works by evolving a population of candidate builds through selection, crossover and mutation. A progress bar under the loading animation tracks generations completed across every worker a run spawns.
A separate single-objective run identifies the highest-damage skill allocation for the highest gear tier, which becomes the gold-bordered "highest damage" card appended to Soldier/Profiteer results.
How daily net cost is calculated
Net cost = gear durability cost + consumables (ammo, food, pill) − scrap resale from used gear − expected case income − battle loot − company engine income − daily base income. All gear cost figures are durability-weighted: a piece that lasts two days costs half as much per day as one that breaks. The "net cost per 1K damage" efficiency figure divides net cost by daily damage, and acts as our primary comparison metric for Soldier/Profiteer builds.
Each income source is its own toggle under Cost Factors and only counts when switched on: Cases and Scrap from combat, Companies (Automated Engine income), and Battle loot (off, a cautious single-battle ceiling, or an average-day estimate). A flat daily base income is always added. Employee (NPC) income is not part of the net cost; with an API token the imported company cards show it separately (see below).
Companies & Prestige
Each company slot holds an Automated Engine tier (1–7) or is marked Inactive, which means never use this company. Importing a profile brings in your disabled companies as Inactive tiles instead of hiding them, and the icon on each tile shows what it produces.
Companies and Management are searched like the combat skills, out of the same skill-point pool, unless you pin them. The Companies skill caps how many companies are active (2 at level 0, one more per level); a build keeps the best ones its level allows. Management adds 2% to the output of every NPC employee at its active companies, up to +22%. Importing your profile with an API token reads each company's roster, so the optimizer can price both from what your employees really earn net of wages; without a token they earn nothing here and the optimizer leaves them at 0. Pin either with the tile under Pin Skills (Management) or in the Companies section (Companies); X lets the optimizer decide.
Prestige points are a separate budget from regular skill points. Pin them per-skill with the steppers under Pin Skills; unpinned ("X") leaves the optimizer free to decide. Multiple skills can stack prestige levels, and cost rises as a power of the number of skills you've prestiged (each additional stacked skill roughly doubles the point cost of the next one).
Corners cut for performance (and why results vary slightly between runs)
- Gear is evaluated per tier, not per individual roll. During optimization each tier is assigned a single representative stat profile (the market value-quality stats, or midpoint if data is sparse). The optimizer cannot see individual rolls — it only chooses tiers. The per-roll recommendation shown in tooltips is computed afterwards, not during search.
- NSGA-II is heuristic, not exhaustive. It searches a sample of the build space, so re-runs with different random seeds may find marginally different results. Results are deterministic for the max-damage pass but vary for the main runs.
- Skill tables are cached per gear combination. Because skill marginal values depend on gear baseline, the optimizer pre-computes skill contribution tables once per unique gear set seen and reuses them. This is the main speed optimization.
- Market analysis uses sampled transactions. Gear stat ranges, value-quality points, and price curves are derived from up to 500 recent market transactions per item (filtered for outliers).
Gear recommendations — Optimal vs Value
Each gear tooltip shows up to three rows:
- Roll range — the full min–max stat range for that slot and tier, from game data.
- Optimal (most efficient for build) — the specific stat roll that maximises total-build-damage divided by price, evaluated against your exact build context (skills, other gear, ammo, food). This is build-aware: it can in principle differ between cards. The price shown is the expected market price at that quality level, interpolated from the observed transaction price curve.
- Value (best stat to price ratio) — the knee of the price-vs-quality curve for this item: the quality percentile at which paying more stops being worth it relative to the stat gain. This is a market-level property, not build-specific.
Why does the same tier show identical gear stats across all build cards? The "Optimal" recommendation is build-aware, but for most armor slots the optimal quality band is robust — the damage-per-price argmax lands on the same point whether the overall build costs 0.15 or 0.25 per 1K damage, because the marginal contribution of, say, pants armor to total damage is small relative to the price curve shape. The 0.15-vs-0.25 difference between cards comes from choosing different tier combinations and skill distributions, not from buying different rolls of the same item. Weapons are the exception — the Optimal recommendation there sweeps a 5x5 ATK/CritC quality grid and is more sensitive to build context.
Prices and live data
The tool needs a Warera API token to fetch live market prices, stat ranges, and transaction history. Without one it falls back to built-in defaults, which may be stale. Token is stored locally in your browser and is never sent anywhere except the Warera API. If you see "Default prices" in the header, click the token prompt that appears after optimizing to enter yours.