Endgame Insights

Shiyu includes complete S-rated teams. Deadly Assault includes complete Normal teams with three stars and Adversity teams with at least one star. Each cycle retains its own observations and frozen Performance tiers.

Teams containing an Agent divided by all included teams in the encounter. All Encounters pools accepted results across the three principal encounters. Integrated is separate. This is a share of team observations, not unique players or owners.

Mean team score associated with an Agent’s appearances, not isolated damage or contribution. Combined views pool all accepted scores, including small encounter fragments. Missing scores are not invented. Deadly Assault combines Trial bosses only. Select Adversity separately.

40% usage percentile and 60% performance relative to other results in the encounter. Usage compares Agents eligible for the period, not account ownership. All Encounters uses the selected pooled population. Unavailable Performance means Unranked Meta Index. This is separate from Team Rankings’ Team Index.

ALL includes every qualifying team. M0 requires known Mindscapes for all three Agents and M0 on every limited S-rank. Standard S-ranks, free S-ranks and A-ranks are unrestricted; teams without limited S-ranks can qualify. Unknown levels are excluded, and M0 never substitutes ALL results.

Performance needs at least 20 distinct contributors in the selected population and a descriptive uncertainty half-width within 10% of its mean. No boss-majority rule applies. Valid observed Usage, including rare or zero appearances, is published independently. Neither metric fills missing scores or requires a tier quota.

Usage bands: S ≥20%, A 10%–<20%, B 3%–<10%, C 0.5%–<3%, D <0.5%. Performance compares eligible Agent averages: S exceptional, A strong, B typical, C below-typical, D weakest supported. C/D do not mean failed clears. These are AstraVoy tiers, not game grades.

Point boundaries are calibrated separately by mode, encounter and investment group, then frozen for this period. Changing filters selects a reference; it does not rebalance tiers. Shiyu’s 25,000-point clear and Trial’s 20,000-point three-star result are gameplay context, not relative tier cutoffs. Meta Index retains its separate 0–100 model: S ≥65, A 55–<65, B 45–<55, C 30–<45, D <30.

Scores are community-reported, not independently verified or representative of every player. Builds, investment and team selection affect results. Separate build and outlier filters from the source are not applied here. Sample precision cannot remove that selection bias. Partial encounter participation can affect comparisons.