Stewarding
For stewards: working the report queue, deciding reports, and releasing flagged sessions.
The steward desk is where the platform's stewards decide the reports drivers could not settle themselves, and check result uploads that failed an automatic check.
Who can steward
Every community stewards its own races. Its owners, admins and members with the steward role see the reports waiting for them under Stewarding on the Organise page, and decide them on the report page as described below.
The platform's stewards are a role the bdr team assigns. They see Stewarding in the top bar, which opens the Steward desk. Everyone else who opens the desk is sent back to the home page. Platform admins also reach the desk from the Stewarding tab of the Platform pages.
You never judge a report you are part of: on a report where you are the reporter or the accused, the decision panel does not appear.
The report queues
The desk lists reports with the status With the stewards, Under review or Appealed, oldest first, in three cards:
- Official races: reports from official communities' races, and any report the platform has taken up. These are yours.
- Appeals: drivers asking you to look at a community's decision once more. Open one, write your reasoning, and select Decision stands or Overturn. Overturning takes back the time penalty, the ban and the Safety Rating penalty.
- Waiting in communities: reports the communities decide themselves, listed so none waits for good. A report waiting 3 days or more is marked. Taking or deciding one makes it the platform's, and your decision final.
A report reaches the queue when:
- the accused selected Contest,
- the accused did not answer within 48 hours, or
- the reason does not allow self-acceptance (for example intentional collision or insults).
Select a row to open the report. See Reports for the driver's side.
Deciding a report
- Open the report. Read the description, the Statement of the accused if there is one, and select Watch the evidence for the video.
- Select Take this report so other stewards see that it is Under review. This is optional: you can decide a report straight from With the stewards.
- In Steward decision, set the penalty. Leave Seconds, SR and Ban days empty to apply the tariff; the placeholders show the tariff values. Any number you enter replaces the tariff value.
- Write your reasoning in the notes field. It is shown to both drivers.
- Select Uphold or Reject.
The incident refund for the reporter always follows the tariff.
What a decision changes
| Uphold | Reject | |
|---|---|---|
| Time penalty added to the accused's race time, race re-classified | Yes | No |
| SR penalty for the accused (races that count for Safety Rating) | Yes | No |
| Incident refund for the reporter | Yes | No |
| Ban, if Ban days is more than 0 | Yes | No |
| Counts towards the reporter's lockout | No | Yes |
After an upheld report, bdr recalculates every rating by replaying all rated races in order. Positions, season points, ELO and Safety Rating of everyone in that race can move, and the change carries through to every later race. The same happens when the accused accepts a penalty. Ratings are also replayed every night.
A ban starts when you decide and lasts the given number of days. While it runs, the driver cannot sign up for races.
Warning
A decision is final in the app. There is no button to reopen a report, change a decision or lift a ban, so check the numbers before you select Uphold.
Contact patterns
bdr looks at every rated race, including pickup races on open servers, for drivers who keep running into other drivers. The contact detection usually cannot tell who hit whom, only that two cars touched. So the evidence is the spread: one driver in contact with many different drivers over many laps.
Every driver with contacts against at least 2 drivers on at least 2 laps gets a certainty between 0 and 100 %:
| Evidence | Adds |
|---|---|
| Different drivers they touched | up to 30 %, full at 6 |
| Different laps with contact (one big crash counts once) | up to 30 %, full at 6 |
| In at least 30 % of the race's contacts | 10 % |
| Waiting laps: laps at least 1.5 times their own median lap, not counting the first | 10 % each, up to 30 % |
| Contacts the detection could tell they caused, minus those they suffered | 10 % each, from −20 % to +20 % |
| Drivers who reported them for that race (rejected reports don't count) | 10 % each, up to 30 % |
Contact patterns lists drivers from 50 %, most certain first: for example 4 drivers on 4 laps and a third of the race's contacts, or 3 drivers on 3 laps plus waiting laps or reports. Reports alone never list anyone. A new report, or a rejected one, updates the race's certainty straight away.
A slow driver everyone runs into has the same spread, but drives steady laps rather than waiting laps. Look at the replay before you confirm.
The columns show the certainty, the different drivers and laps, the driver's share of the race's contacts, their waiting laps, reports, contacts caused and suffered, and their earlier findings of the last 90 days. The same pattern across several races is much stronger evidence than one race.
- Confirm: the driver takes the incident points of all their contacts in that race, and the drivers they touched take none. Ratings are recalculated.
- Dismiss: nothing changes.
A finding does nothing until you confirm it, and only stewards see it. Latest verdicts lists your last decisions. Undo puts one back in the list, and undoing a confirmation recalculates the ratings again.
Note
The detection works on positions and speeds sampled a few times per second, and it can be wrong. Check the replay before you confirm.
Flagged sessions
When a result upload does not match what bdr expected, the session is flagged and stays unrated. Flagged sessions lists them with Received, Host, Race and Findings:
| Finding | Meaning |
|---|---|
track_mismatch |
The result is from a different track than the race |
server_mismatch |
The result came from a server the split was not assigned to |
unknown_driver:car_N |
Car N was driven by someone without a seat in that split |
To release a session:
- Open the race from the Race column and compare the result with what should have happened.
- If the result is genuine, select Release and confirm.
The session is rated shortly after and moves ratings and points like any other race. A session you do not release stays unrated.