Online Stats Analysis (Game)
Online Stats Analysis (OnlineStats) estimates PCU (Peak Concurrent Users) / ACU (Average Concurrent Users) from session intervals, measuring the game's concurrent capacity and popularity. It corresponds to the backend AnalyticsService.OnlineStats.
Game-specific model. It can be filtered by game server (
server_id) and depends on session data.
1. What Questions It Answers
- What is the PCU (peak concurrent users) within the statistics period? When did it occur?
- What is the ACU (average concurrent users)? How many times the average is the peak? (concurrency fluctuation)
- What is the concurrent online scale of a given game server?
2. Key Metrics
| Metric | Meaning |
|---|---|
| PCU | Peak Concurrent Users, the maximum concurrent online count within the period |
| ACU | Average Concurrent Users, average concurrent online count |
| Statistics duration | duration_minutes (minutes) |
| Total sessions | total_sessions |
3. Backend API
gRPC: AnalyticsService.OnlineStats
| Field | Type | Description |
|---|---|---|
time_range | TimeRange | Analysis time range |
app_id | uint32 (optional) | Filter by app |
server_id | string (optional) | Game server filter |
Response OnlineStatsResponse: pcu / acu / duration_minutes / total_sessions.
HTTP (admin proxy)
POST /admin/v1/analytics/online-stats
Content-Type: application/json
{
"timeRange": { "startMs": 1718169600000, "endMs": 1718774399000 },
"serverId": "5"
}
4. Implementation
PCU/ACU are estimated from session intervals, not from real-time heartbeat counts:
- Each session has a start/end time, forming an "online interval".
- All session intervals are overlaid on the timeline; the number of overlapping intervals at any moment is the concurrent online count at that moment.
- PCU = the maximum concurrency within the period; ACU = the average concurrency.
PCU/ACU are therefore estimates; their accuracy depends on whether session starts/ends are recorded accurately.
5. Typical Scenarios
Server capacity planning
Look at historical PCU peaks to plan server concurrency capacity, ensuring peak periods (launches, events) do not crash.
Event popularity evaluation
Query PCU/ACU during an event period and compare with normal levels to evaluate how the event lifted online popularity.
Server popularity comparison
Query by server_id separately and compare concurrent scale across servers to identify popular and dead servers.
6. Notes and Caveats
- Estimated, not real-time: PCU/ACU are estimated from session intervals and may differ from real-time heartbeat counts; for precise real-time online numbers, use dedicated heartbeat reporting.
- Depends on session data: session start/end times must be accurate; the timeout splitting rule affects results.
- Time range impact: PCU is the peak within
time_range; a wider range is more likely to include extreme peaks. server_idfilter: games can view online stats per server, which requires the C# SDK to reportserverId.