Sentinel Network computes reputation scores for every device based on its ban history. Scores indicate the likelihood that a device is problematic — higher scores mean higher risk.
Reputation scores are calculated based on the number of active bans of a given type. Each additional active ban increases the score rapidly at first, then approaches but never reaches 100%.
Score Examples
| Active Bans | Score | Interpretation |
|---|
| 0 | 0.0% | Clean record |
| 1 | 50.3% | One active ban — moderate risk |
| 2 | 75.3% | Two active bans — high risk |
| 3 | 87.8% | Three active bans — very high risk |
| 5 | 97.0% | Five active bans — near-maximum risk |
| 10 | 99.9% | Approaching 100% asymptotically |
The formula ensures:
- A single ban immediately produces a meaningful score (~50%)
- Scores increase rapidly with additional bans
- Scores asymptotically approach 100% but never reach it
- Revoking bans reduces the score (only active bans count)
Score Types
Each device has two independent reputation scores:
| Score | Based On | Description |
|---|
cheat_score | Active cheat bans | Risk of gameplay manipulation |
social_score | Active social bans | Risk of community violations |
Scope Aggregation
Reputation is computed at two levels:
Global Reputation
Counts only global-scope bans. Represents the device’s network-wide trust level.
Publisher Reputation
Counts publisher and game-scope bans issued by a specific publisher. Represents trust within one publisher’s ecosystem.
Device Check Response
The API returns the higher of the publisher-level and global-level scores (based on your policy settings). This ensures a device with a poor global reputation cannot appear clean to a publisher that has opted into global reputation visibility.
Publisher Control
Publishers control which reputation scopes contribute to their device check responses via policy settings:
| Policy Flag | Effect |
|---|
rep_include_game | Include game-level bans in reputation |
rep_include_publisher | Include publisher-level bans in reputation |
rep_include_global | Include global-level bans in reputation |
Even if a publisher opts out of global reputation, global bans still exist and are visible to publishers that do include them. Opting out only affects the scores returned in device check responses for that publisher.
Reputation Updates
Reputation scores are not updated synchronously during ban operations. There is a brief processing delay — typically under 1 second — before a ban is reflected in reputation scores. Reputation data is also cached, so device check responses may return slightly stale scores for up to 60 seconds after a ban operation.
For safety-critical decisions immediately after issuing a ban, query the ban list directly using GET /v1/device/{device_id}/bans/{type} rather than relying on the reputation score in the device check response.