Balance Lab
Repeatable experiments, using the game’s combat engine. Experimental settings never change the playable game.
Choose a test
Edit equipment, attributes, perks and growth below. Perks map catalog IDs to ranks 0–3; growth maps attribute IDs to nonnegative earned points beyond creation. Example: "perks":{"counter":3,"storage":2}, "growth":{"strength":8}. Spend exactly 10 attribute points, or omit attributes to use the baseline. The opponent selector above is ignored for this test.
Trials alternate fighter positions. All styles stay fixed. Samples allocate points sequentially at random and choose equipment independently; repeated builds are possible. Field results use this selected field, not every possible character. Matched seeds preserve starting randomness; later rolls can diverge when combat paths change. Attribute category rates are associations, not controlled point-transfer effects.
Origins and sampling
Every equipment combination includes all six origins. Sampled builds choose origins uniformly with equipment, independently of allocated points. The fixed trainer uses Initiator. Custom builds accept an origin ID; omission retains legacy no-origin behavior. Origin modifiers apply after allocation and growth. Numeric origin rules are editable below, without changing playable defaults. Metrics count both fighters, including prepared and consumed effects; these are not causal win-rate estimates. Origin descriptions show default rules. Old saved runs retain their engine snapshots and are historical.
Weapon catalog and effects
Equipment and sampled fields include all 12 weapon variants. Tournament creation remains basic-only. Category perks apply to all variants. Weapons and perks add percentage damage bonuses; style and Strength multiply separately. Weapon rules are independently editable below. Effect metrics count both fighters and distinguish prepared attempts from landed damage; they are not causal estimates of win-rate impact.
Perk catalog and sampling
Progressed samples cycle across perk IDs for coverage, randomly assign ranks and additional perks, and add random attribute growth. They are stress-test builds, not a model of natural reward frequency. Category associations are not causal effects. Counterattacks, their damage, absorbed damage and peak stored armor damage are recorded. Perk descriptions show playable defaults; experimental numeric overrides are recorded in each run.
Experimental rules
Start from the current rules, then change numbers to compare. With no edits, both columns should match exactly. Probability values use 0–1; defense modifiers use percentage points. Reckless Power uses recklessDefense for rank 1, recklessDefenseRank2 and recklessDefenseRank3 for the later ranks, and recklessSelfDefense for its own defensive penalty. Dagger Specialist uses dagger for rank 1, daggerRank2 for rank 2, and daggerRank3 for rank 3. Opening Gambit experimental controls: openingMomentum adds opening momentum per rank; openingDefensePenalty subtracts fixed defense points whenever its owner defends through openingEnd. Defaults are 5 opening momentum per rank and a fixed 10-point defense penalty. Becoming defender triggers the drawback, including round 1; it affects all remaining defenses in that window. Gambit metrics count both fighters and report actual defense rolls changed by the penalty, not duel wins. Dodge Preparation gains use dodgeStep, dodgeStepRank2 and dodgeStepRank3 (5/7/10); rank 3 exceptional dodges double its gain. These controls are independent; the catalog describes playable defaults.
Weapon, armor, style, and combat settings
Ready.
Results
Saved runs
Last five completed runs are kept in this browser when storage allows. Export important runs. Historical results retain their original rules and engine identity.