PurrwikimaniaA Sword x Staff reference
English

Combat mechanics

Raid boss simulator

How much damage does your party deal to a boss? Pick the Chaos Tyrant raid or one of the Crucible of Hieromon’s bosses, put one to four players in the party, from the top-100 list or typed in from your own stat screen, set the boss’s tier or its level and rank, and run a few hundred fights on the same engine as the PvP simulators, with the game’s PvE rules. You get the average damage and actions, a chart of the damage each Technique and Charm deals, and the fight closest to the average to watch.

PvP 1v1 2v2 4v4

The Crucible changes its boss every week. Picking another boss loads its own field.

Chaos Tyrant — 5 × 5the fight’s own field; Y up, as the game counts
Your party — the 3 × 3 boxdrag a player (or tap it, then a cell) inside the box

Your party one to four players

The boss

Combat log click a line to jump there

    Top 100 by combat rating

    Drag a row into the party (or tap a row, then a slot). This week’s Tournament bracket, played on this engine: Tournament projection.

    Add your own fighters

    Copy your character, or any player’s Player Info, from the game: the stats list as it prints it, your class, level and rank, the four Techniques and four Charms, your Fantomon, and the gear lines that act in a fight. Fighters you add stay in this browser, at the top of the list, marked yours. How to fill it in.

      the rank and level this slot is equipped at; a change applies to the slot at once

      The look is for the battle view and stays in this browser. Names and icons are the game’s.
      Pick the Fantomon, then its form, level and skills below.

      How the simulator works

      Checked against a recorded raid

      One solo run was recorded on 15 November 2026 and compared with the simulator hit by hit. The fight ran the full 25 rounds and the boss took 17 turns, in both. Single hits matched the server to the last digit (Wind Blade Spiral: 10,584,416 in the simulator, 10,584,408 in the game). Over 100 simulated runs the same setup dealt 3.21 billion on average, with 8 in 10 runs between 3.10 and 3.32 billion. The recorded run dealt 3.14 billion, and an earlier run with the same setup 2.96 billion.

      Not modelled

      Where the numbers come from

      The raid’s rules come from the game’s own files: the 25-round cap, the field, the 3 by 3 box the party sets up in, the boss’s body, stats, Techniques and passives, and the PvE damage rules with their level and rank gap tables. The boss’s starting cell and the order it casts in come from the recorded run. One rule is an inference from that run: a Technique aimed at the boss lands around the boss’s middle, not around the cell it was aimed at. Divine Wrath aimed at the boss’s edge struck it 13.4 times in 16, the share an area centred on the boss gets. The 25-round end, the 17 boss turns and the single-hit damage were checked against the same run.

      The Crucible of Hieromon’s bosses come from the same files: each one’s field, body, Techniques and passives, the 10-round limit, and the level and rank of every tier. A boss stands where the stage’s formation puts it from the party’s entry point, the rule the raid’s recorded starting cell fits. One Crucible stat screen was checked: the Storm Griffin at Master tier 1 matches stat for stat. The simulator plays these parts of a Crucible boss: its Techniques, the Rage at half HP with the skills it unlocks and the passives it swaps, the strike it answers with for every 15% of its HP it loses, the stat boost for every 5% it loses, and the element it resists and the one it is weak to. Where the game applies the element rule in its formula is not known, so it is taken as a plain multiplier on the hit. It does not play the things a boss summons (the Griffin’s orbs, the Orochi’s Fishlings), the Primordial Loong’s change of form, or Yoldra’s shift between Frostchill and Blazing. No Crucible fight has been checked against a recording yet.

      The loadout search cannot fight every loadout: a tier 5 class has about 3 billion. It fights a few thousand random ones on the same dice and fits a model to them, with a part for each skill and a part for each pair of skills. The model ranks every loadout, its 48 best are fought for real, and the best of those is checked against every single swap and every cast order. In a test on one recorded player the model was about 4% off on loadouts it had not seen. The answer is the best loadout found, not a proof. Two things are inferences. A class can equip the skills of every class before it on its line: every recorded top-100 loadout fits that, and no tooltip states it. And only the skills seen in a recorded fight have been checked against the game, so a skill the simulator reads wrongly will be ranked wrongly too.