Top-X strength curve
Average strength of the strongest X characters, ranked by Combat Power
SERVER STRENGTH INTELLIGENCE / FINAL DATASET
Compare the power ceiling, roster depth and progression investment of two AION 2 servers. Results are calculated once from the finalized dataset. No live API or database is needed to view this publication.
Fixed analysis criteria: level 45+ · all classes · all item levels · CP over 320,000 excluded. Competitive thresholds: 250,000 CP and 3,500 item level. The data is not updated after export.
HEAD-TO-HEAD
Average strength of the strongest X characters, ranked by Combat Power
Equivalent fixed-size rosters — exact sample counts
| Top X | Server A avg CP | Server B avg CP | Difference |
|---|
A top-X average appears only when X indexed characters have valid CP.
Same CP bins across both servers · indexed character counts
How many discovered profiles reach the power targets?
These are indexed-character counts, not verified currently-online populations.
Compare the middle and the upper tail of both server populations
RELATIONSHIPS
Frozen character-level scatterplot, with one fitted line per server
Pearson r and trend lines use every eligible character. The scatterplot samples at most 750 points per server; the density plot bins ALL eligible profiles into 48 × 32 cells per server. Correlation does not establish causation.
Item-level brackets of 250; compare Combat Power and progression within each bracket
CHARACTER DEVELOPMENT
Average unlocked nodes per reported board
Total nodes unlocked is not the same as Daevanion points spent. Points spent require individual board-node cost data. Missing boards are excluded, not counted as zero.
Ready to analyze when equipment skill data and a verified skill-level-to-point formula are added. Existing character-info snapshots do not expose those skill levels.
COMPOSITION
Numbers and mean strength by class, server against server
NAME-DISCOVERY COHORTS
Character names and the writing systems they contain — not a player's nationality, ethnicity, primary language, prior game experience or spending. Han characters are shared by Chinese and Japanese. Hangul suggests a Korean writing system; kana suggests Japanese writing. Mixed names can appear in multiple script-presence categories.
Two different denominators: indexed shares count all discovered names meeting the minimum-level filter; class, item-level, CP, and Azphel filters apply only to collected profiles. Top-X representation ranks those valid profiles by CP. An origin labeled unknown cannot be assumed to be Latin; unknown discovery origins remain unknown.
Latin-only, Latin mixed with non-Latin, non-Latin-only, and numeric-only names. Every name belongs to exactly one row.
Hangul, Han, kana, Cyrillic, and other writing systems. Rows overlap for names using multiple scripts; these are not population totals by nationality or language.
Where the character ID was first recorded: Latin, Unicode, or numeric searches. Historical origins come from archived search pages, where recoverable.
Share of CP-ranked top X versus share of all harvested profiles with CP. Ratios over 1.0 mean that cohort appears more often at the top than in the valid-CP profile pool.
Top-X results are suppressed until each server has at least X valid CP profiles. Profile selection bias remains possible.
METHODOLOGY
Indexed characters are not the server's entire active population. Discovery coverage may have gaps, and a level-45 profile does not establish recent login activity. Compare servers harvested at similar times, and treat unmatched harvesting coverage as a potential source of bias. Numeric averages exclude missing values; sample counts are displayed separately.
Scaled PvP contamination: Profiles reporting over 320,000 CP are excluded from all analysis, including CP, item-level and Daevanion averages, top-X rankings, name-script cohorts, scatterplots and character tables. Excluded counts appear in Data Quality. Historical raw responses remain private and are not included in this static export.
PvP military rank remains unavailable from the tested Global API. No unknown rank is treated as unranked, and the published calculations never assign zero to missing Daevanion or stigma values.
RAW COHORT EXPLORER