Choose the Keqing Example or start My Setup. Roster, tests, and simulator settings stay in this browser. An invited account can keep a copy too.
Each person can tailor the roster and simulator on their device. Setup downloads include the roster, personal tests, simulator settings, and saved simulation history.
Team DPS is damage per second for all four characters in a linked setup. Fresh reruns used the official gcsim v2.47.6 CLI on existing public configs; published simulations came from the gcsim database; reported dummy tests came from the original spreadsheet. None predicts your account’s damage.
Open Builds & test conditions to check constellations, weapons, artifacts, targets, and duration. A simulation’s full duration is not necessarily one rotation. Different methods and investment levels are not controlled comparisons.
A fresh rerun repeats an already counted public build and rotation; it is not independent team evidence. The overview counts each original config once. The new Keqing Lunar-Charged runs use a KQM example rotation, adapted example builds, and explicitly note stamina delays. Character damage shares come directly from gcsim output. Some catalog characters have no collected valid C0 config, so their rerun is unavailable.
KQM supplies team examples. A source number attaches only when all four selected characters match that source composition; its builds still apply. Substitutions recalculate the separate model range and only gain source DPS when an exact composition is documented. Guide order is not a power ranking.
Model range: a statistical guess from published one-target gcsim results with all 5★ teammates at C0. It uses character and teammate-pair patterns; it does not run a new simulation for the displayed lineup. 4★ constellations, weapons, artifacts, rotations, and buffs vary across the sample. Unsupported characters and flexible slots get no estimate.
The model trained on 1,373 simulations after removing results below 10k or above 300k DPS. In five-fold checks holding out exact four-character compositions, median relative error was 7.9%. Similar teams dominate that check; unfamiliar teams can miss by much more. The displayed ranges widen as matching evidence gets thinner. They are planning guides, not a statistical guarantee.
The Simulation lab preserves the validated Keqing / Columbina / Sucrose / Ineffa runner and exact source presets. Supported characters can use live edits to level, constellation, base talents, compatible weapons, artifact sets, main stats, and artifact rolls. The three source builds whose substats cannot be converted exactly keep their other live controls. Live results, character contributions, and comparisons come from gcsim.
The team builder starts from source builds, lets you swap teammates, and composes a generated rotation for supported dynamic teams. Exact presets retain their exact source rotations. The weapon picker uses the current gcsim catalog; the artifact picker uses supported sets, including 4-piece and 2-piece combinations. Custom rotations and test conditions remain preview controls until those paths are supported by live gcsim. Generated rotations are exploratory, and unsupported character mechanics or missing build evidence produce an explicit unavailable state.
gcsim simulation database ↗
Original test sheet ↗KQM Keqing guide ↗
Invite-only. Use the email that received an invitation. There is no public signup.
This device and your saved account have different setups. Sync waits until you choose one.
This restores the Keqing Example on this device.