Freeze the source set
Record the lineup, source URLs, access time, festival edition, and every known stage or host.
PUBLIC MODEL CONTRACT · VERSION 1THE METHOD
THE METHOD
BEHIND THE CALL.
Profile the stages first, score every viable artist-stage pairing, preserve the runner-up, then lock the forecast before official assignments are known.
01 · THE PIPELINE
FROM PUBLIC LINEUP TO LOCKED RECORD.
The same sequence applies to every festival, while each edition gets its own evidence and stage profiles.
Profile stages before artists
Define each stage by sound, scale, curators, historical booking patterns, capacity, and special roles.
Build artist evidence
Capture same-festival history, style, billing, production scale, special billing, trajectory, and related-event placements.
Score every viable pairing
Apply the 100-point rubric to each artist-stage candidate. Missing evidence is neutral; contradictory evidence must reduce the score.
Resolve the whole lineup
Apply hard constraints and balance the full day. Keep the winning stage, runner-up, score gap, rationale, and evidence trail.
Assign calibrated confidence
Use evidence strength, candidate margin, and unresolved conflicts. Confidence is not the same as the raw fit score.
Lock, publish, and wait
Timestamp the immutable forecast and methodology version. Official information is stored later without overwriting the call.
Score and learn
Award one point for an exact stage match, zero for a miss, exclude cancellations, then publish accuracy and calibration.
02 · CANDIDATE SCORE
ONE RUBRIC. EVERY VIABLE STAGE.
The fit score totals 100 points. It ranks candidates; it does not claim to be the probability of a correct prediction.
30PTSSame-festival history
Recent exact-stage appearances at the same festival carry the strongest signal.
25PTSSound + programming fit
Genre, set style, stage identity, and the way the festival programs that room.
15PTSBilling + production scale
Artist billing, audience demand, show scale, and likely placement in the day.
10PTSHost + curator fit
Announced stage hosts, labels, collectives, or curated programming restrictions.
8PTSSpecial set signals
Sunset, live, B2B, takeover, debut, or other explicitly billed formats.
7PTSEvent-family evidence
Placements at related events are supporting evidence, never a replacement for local history.
5PTSOperational balance
Capacity, day balance, competing headliners, local stages, and schedule feasibility.
THE WHOLE LINEUP MATTERS.
A top individual score can still be impossible in context. Before locking, the model applies these edition-wide constraints:
- Announced host, label, or stage-eligibility restrictions
- Festival day and artist billing identity, including B2B sets as one billing
- Mutually exclusive special slots such as sunset or closing windows
- Capacity and headliner balance across the complete day
- Local or sponsored-stage eligibility when publicly identifiable
- An explicit unknown-stage option when the public stage inventory is incomplete
EVIDENCE + MARGIN, NOT INSTINCT.
90–95%Direct precedent
Recent same-festival exact-stage evidence with a decisive score gap and no unresolved conflict.
80–89%Strong convergence
Several independent signals point to the same stage.
70–79%Clear fit
The best stage fit is clear, but direct local precedent is limited.
60–69%Competitive call
The leading stage narrowly beats a credible adjacent-stage option.
50–59%Low evidence
Local, emerging, sponsored-stage, or otherwise thin-evidence placement.
03 · IMMUTABLE RECORD
LOCK FIRST. SCORE LATER.
The prediction and the answer live in separate records, so hindsight cannot improve the forecast.
Saved at forecast lock
Festival and edition IDFramework versionSource URLs and timestampsStage profilesArtist evidenceCandidate scoresWinning stage and runner-upScore gapStage confidenceTiming forecast and separate timing confidenceRationaleForecast lock time
FOUNDING MODEL NOTE
The 2026 Orlando forecast predates both this formal rubric and the published exact-match scoring contract. Its 108 calls remain unchanged and are labeled as the founding legacy record. The scoring rule was locked before official results, but not before the forecast; every future edition must lock both methodology and scoring prospectively.
View the 2026 Orlando archive →04 · CALIBRATION REPORTING
EVERY RATE NEEDS ITS SAMPLE.
Observed accuracy must be shown with its scored sample size and uncertainty; a small bin is evidence in progress, not a stable performance claim.
0Unscored
No eligible result is available.
1–9Preliminary
Too few eligible calls for a stable calibration conclusion.
10–29Developing
Useful directional evidence; uncertainty remains material.
30+Established
Large enough for routine comparison while retaining the uncertainty interval.
Observed accuracy uses a 95% Wilson score interval. Calibration error and Brier score remain model-version specific; bins are never merged across incompatible prediction standards.
05 · LISTENING COMPANIONS
ONE LINEUP. ONE CONSISTENT PLAYLIST METHOD.
Every released festival lineup receives the same reconciled, private-first playlist workflow. Collection may be automated; public presentation requires an exact owner-approved Spotify record.
Verify the release
Confirm the official lineup through registered authoritative sources and approve the material monitoring change.
Freeze the artist source
Save the approved lineup snapshot, source URL, checked time, source order, and unique billing identities.
Generate the private manifest
The released-lineup handoff automatically creates a hash-bound, non-public playlist draft.
Reconcile Spotify identities
Match every billing to the correct Spotify artist and record ambiguity or missing-catalog exceptions.
Select representative tracks
Apply the declared coverage and track-selection rules consistently across the complete lineup.
Create privately
Use the authorized Spotify integration to create the playlist privately in the connected account.
Review exact output
Check coverage, duplicates, identity matches, title, description, artwork, ordering, and exceptions.
Approve and make public
Record exact owner approval, then change the approved playlist to public in Spotify.
Register and embed
Store the exact public Spotify and embed links; the shared festival component renders the companion.
Maintain an audit trail
Treat later lineup or catalog changes as revisions with evidence, review, and approval before publication.
Coverage standard: one representative official-catalog track per unique billed act when available. Identity ambiguities and missing catalogs are recorded as exceptions, never silently substituted. Spotify creation is private-first; public embedding requires exact owner approval.
06 · SEALED DEFINITIONS
THE RULES HAVE RECEIPTS, TOO.
Each published prediction, scoring, calibration-reporting, playlist-publication, and sanitized-workbook definition has a canonical SHA-256 fingerprint. A definition cannot change under the same version: changes must be published as a new version.
Prediction Standard v1
sha256:f64a6f67604d1bfd133c03cb819e0e661777915326cd561f889d1f0a4ebc5ac9
Calibration Reporting v1
sha256:768addbb8ee6c5361aae0bb3475b8613f74309f5f8adee3586a591351fac4eb5
Listening Companion Standard v1
sha256:f8f34a7b82b20cad40ca29ec1c0f3ef7259a064c0c3ad8c54559f94c7a5c1682
Scoring Contract v1
sha256:56e54f524db0e5c2651a46f9c4ad561d8290c3c7b09d86b0e4375ae5c33f294e
Public Workbook Template v1
sha256:ae2b91f15d1ea4a562b900feb0169dce899a4fe80d5ad5c7fd45bc5cd12048e1
Download the machine-readable methodology →Read the independent brand policy →