Audit competitions
How contests run, how findings are judged blind, and exactly how the prize pool is split.
An audit competition is a time-boxed review of a frozen commit with a fixed prize pool. Anyone can compete, and the scoring formula is published before day one.
Phases
| Phase | What happens |
|---|---|
| Upcoming | Scope, rules and scoring are published and locked (fingerprinted) |
| Live | Submit findings until the end time |
| Judging | An independent judge reviews every finding |
| Escalation window | Preliminary results are published; 48 hours to escalate |
| Finalized | Final standings and payouts |
Submitting a finding
Give a title, a severity (High or Medium, per the contest's criteria), the location (file#Lx-Ly), the details, the impact, a proof of concept, and optionally a mitigation. Findings are encrypted and visible only to you and the judge. At finalization, group titles, severities and the payout table are published. Full write-ups follow the contest's disclosure policy.
Blind judging
The judge sees each warden as a pseudonym (e.g. W-9765ab), never your name, so reputation can't bias a decision. Every decision (valid, invalid, severity, duplicate grouping) includes a written reason you can read.
Duplicate groups
Valid findings are grouped by root cause into labels like H-01 or M-02. You can see your group's size and severity and your share of it.
Scoring
value(group) = points[severity] × 0.9^(n − 1) n = findings in the group
share = value(group) ÷ n
best write-up = share × 1.3 selected by the judge
your payout = prize_pool × your_points ÷ total_points
Defaults: High = 10 points, Medium = 3 points. A unique finding is always worth more than the same finding shared. None of the pool is paid at the platform's discretion, and the rounding remainder is shown and returned to the sponsor. Anyone can recompute any payout from the public results.
Escalations
When preliminary results are published, you have 48 hours to escalate each judged finding once. Cite the rule or severity criterion the judgment conflicts with. The judge must respond with reasoning (uphold or revise) before results are finalized.