Methodology and privacy
Research pages are designed around aggregation, minimum samples, coarse operational labels and source-specific definitions. They never expose an individual practice record.
Practice records
Queries use solve time, timer mode, stage timestamps, move count, trainer components and event completion only when needed for the named measure. Public outputs contain grouped rows, never scrambles, move sequences, account names or individual solve histories.
Small groups
SupaKewber benchmark cells require at least 20 distinct users and 100 observations. Anonymous connection telemetry has no account identifier and requires at least 100 outcomes. Below-threshold pages remain “collecting.”
Bluetooth operations
Connection events store only success, failure or unexpected disconnect; coarse browser, operating-system and cube family; an approved error bucket; and duration. No IP, account, MAC address, raw user-agent string or persistent fingerprint is stored. Global Privacy Control and Do Not Track disable the event.
Reproducibility
Each JSON response includes sample unit, definition, calculation method, inclusion and exclusion rules, date range, update date and stable download URLs. Material definition changes should create a documented version rather than silently rewriting history.
Official data boundary
WCA-derived pages use the public V2 export, show its export date and format version and reproduce the required attribution. The age analysis remains intentionally unavailable because the public export excludes birth dates.
Corrections
If a query, label or source issue is discovered, SupaKewber can update the calculation while preserving the URL, update date and method. A result should be quoted together with its sample, period and population.