Step 01
Define
An editor writes one resolvable question, fixes the outcome taxonomy, and records the source standard.
Will They Hire?Audition outcome forecasts
Methodology · version 0.9 preview
A forecast is only as goodas its resolution.
Cadenza treats each forecast as a small public research object: defined before the event, locked on time, resolved from evidence, and corrected without erasure.
Method checksum
CF-METHOD-0.9
Five exhaustive outcomes
Pre-registered lock
Human evidence review
No candidate names
Public evidence only
Human-reviewed results
Calibration, not hype
The lifecycle
Five controlled transitions
Forecasts move in one direction. Each transition creates a timestamped record that can be inspected later.
Step 02
Verify
The underlying public listing, schedule, organization, and lock time are checked before publication.
Step 03
Lock
Forecasting closes before the first advertised round. The locked aggregate and every revision are preserved.
Step 04
Resolve
A human reviewer classifies the first qualifying public confirmation and attaches its evidence.
Step 05
Score
Proper scoring compares each probability estimate with the verified outcome and updates calibration.
Forecast object
The contract is frozen first.
Participants see the same question, outcome definitions, time boundary, and evidence standard. Editorial changes create a new visible version.
questionfirst publicly confirmed outcome
locks_atbefore first advertised round
resolutionfirst qualifying public source
identitycandidate data prohibited
versionappend-only history
Outcome taxonomy
Exhaustive without becoming personal
These five outcomes describe what the organization publicly does. They never ask who applied, advanced, or was privately considered.
1
Permanent appointment
The advertised permanent seat is filled.
2
Trial or temporary appointment
A trial, acting, interim, or fixed-term appointment is confirmed first.
3
No appointment
The search concludes without an appointment.
4
Cancelled, postponed, or reopened
The process materially changes before another outcome is confirmed.
5
No public confirmation by deadline
No qualifying public source appears before the resolution deadline.
Proper scoring
Reward honest uncertainty, not bravado.
The multiclass Brier score compares the complete probability vector with the verified outcome. Lower is better. Expressing 65% confidence should mean that similarly rated calls resolve that way about 65% of the time.
BS = Σ (pi − oi)²
pᵢ
Probability assigned to outcome i
oᵢ
1 for the observed outcome; 0 otherwise
Σ
Sum across the five possible outcomes
Worked example
Trial or temporary appointment resolves
Permanent25%
Trial / temporaryobserved55%
No appointment10%
Search changed5%
Unconfirmed5%
Brier score
One resolved example
0.280
Lock integrity
The authoritative server time controls whether a revision is accepted.
Independent first call
The aggregate view remains hidden until a valid forecast is recorded.
One account, one record
Rate limits and account checks protect the aggregate from duplication.
No identity targets
Candidate names, rumored applicants, and private rosters are not accepted.
Corrections
Change without erasure.
A source can be retracted, clarified, or superseded. Corrections append a new classification, reason, reviewer, and timestamp while preserving the prior state.
Version 1 · 14 Jul · 09:10 UTC
Trial or temporary appointment
Version 2 · 14 Jul · 16:42 UTC
Permanent appointment
The organization clarified the employment term in a revised public announcement.
Questions
Method notes
Why use the first public confirmation?
It creates a single observable stopping rule. Without it, later announcements could repeatedly rewrite what the original forecast meant.
What counts as a qualifying source?
An official organization page, public announcement, official roster, or another source explicitly accepted by the forecast contract. Anonymous claims and private messages do not qualify.
Can a participant add a candidate name?
No. Public employment forecasts remain organization-level. The product does not accept candidate names, private applicant lists, or rumors about individuals.
Does leaderboard position create a benefit?
No. The leaderboard is gated design research and is not part of the live pilot. If a future aggregate display ships, it will create no reward, special access, discount, or economic entitlement.
See the method in motion.
Record a local preview forecast and reveal the illustrative aggregate view.