Measuring FOMO before
the crowd feels it.

Each narrative receives a composite score on a 0–100 scale, recomputed every 15 minutes. The score does not measure price or value — it measures attention compressing relative to the narrative's own 90-day baseline. A reading of 60 means attention is accelerating faster than in 60% of comparable historical windows.

Normalization

Each raw input is converted to a z-score against the narrative's own 90-day rolling baseline, then winsorized at ±3σ so a single viral event cannot pin the composite. Φ is the standard normal CDF, mapping the weighted sum onto 0–100. Because baselines are per-narrative, a small narrative and a majors narrative are comparable: both measure acceleration relative to their own history, not absolute size.

Narrative taxonomy

A narrative is a curated basket: 5–40 tokens, a canonical term set for search/social matching, and a contract set for on-chain telemetry. New narratives enter through governance listing with a 14-day probation during which they carry a PROVISIONAL badge and are ineligible for Discover Mining. Basket membership is versioned; every snapshot references the basket version it was computed against.

Update semantics

• Cadence. Recomputed every 15 minutes at :00, :15, :30, :45 UTC; publication lag target < 90 seconds.
• Continuity. Component z-scores are EWMA-smoothed (half-life 45 min) — the score moves in points per interval, never in cliffs, except on verified news events flagged in the snapshot.
• Determinism. Given the same inputs and basket version, any party can recompute the score bit-for-bit; the pipeline is open-source.

The score maps deterministically onto five phases. Phase boundaries are part of the oracle contract: markets, mining multipliers, and warnings all settle against them.

Boundary hysteresis

To prevent flapping at phase edges, a transition only registers after the score holds across the boundary for 2 consecutive snapshots (30 minutes). Markets referencing "enters phase X" settle on the registered transition, not the first touch. The one staged UI moment — the Ignition flare — fires exactly once per registered upward crossing of 20.

Every 15 minutes the oracle publishes a signed on-chain snapshot containing the six component z-scores, the composite score, the phase, and the Organic Ratio for each tracked narrative.

• Settlement. All FOMO Markets and mining evaluations settle exclusively against published snapshots — never against live feeds.
• Disputes. A 2-hour window follows each settlement. Challenges require a 1,000 $FOMO stake, slashed if the snapshot is upheld.
• Degradation. If an input feed fails, the last good snapshot is served with a staleness badge (◔ 15 min ago). Latency transparency is part of the oracle contract, not an apology.

Pipeline stages

• Ingest. Exchange APIs, archive-node indexers, social firehoses and search feeds land in a staging lake with source signatures.
• Filter. The Organic Ratio pass (§04) strips synthetic attention before any component is computed.
• Compute. Deterministic scoring workers produce component z-scores and the composite; three independent workers must agree within ε = 0.01 or the snapshot is withheld.
• Publish. The snapshot is Merkleized, signed by the oracle key quorum (3-of-5), and posted on-chain; the UI reads the chain, not the pipeline.

Snapshot schema

Dispute resolution

• T+0 → T+2h. Anyone may challenge a settlement by staking 1,000 $FOMO and citing the disputed inputs.
• T+2h → T+26h. A recomputation committee re-runs the pipeline from archived raw inputs; the result is binding.
• Outcome. Upheld snapshot → challenger stake is slashed 50% to the insurance pool. Overturned → challenger earns 2× stake and affected settlements are replayed.

The share of measured attention judged human after bot filtering — account-age clustering, posting-cadence entropy, and funding-graph analysis over the mention set.

Filter methodology

The ratio is recomputed per snapshot over the trailing 6-hour mention set. Filter parameters are published one version behind (n−1) to resist adversarial adaptation while remaining auditable.

Reputation — the Hunter Score

A soulbound composite: Early Detection 35% · Prediction 35% · Signal 20% · Tenure 10%. It cannot be bought, transferred, or delegated; B2B signal subscriptions are priced against it.

Protocol constants

27 completed narrative cycles, October 2024 – April 2026. Full per-cycle data ships quarterly with the oracle snapshots.

Method

Walk-forward evaluation with no lookahead: each cycle is scored using only the specification version live at the time, against baskets frozen at cycle start. "Peak price volume" is the 24-hour window with maximum basket dollar volume within ±60 days of the Ignition crossing.

Sample cycles

$FOMO is the protocol's work token: every earning mechanism requires staking it, and every reward is denominated in it.

• Sinks. 0.1% of every market fill is burned; negative Brier scores burn; manipulation slashes burn 50% (the rest funds insurance).
• B2B flywheel. Institutions subscribe to hunter signals in $FOMO; 70% streams to the hunter, 30% to the treasury.
• No emissions for holding. $FOMO pays for being early and right — never for being idle.

• Oracle inputs depend on third-party data feeds; degraded feeds widen settlement uncertainty.
• Phase boundaries are specification constants — they are not tuned to any individual narrative.
• Reputation slashing for manipulation is irreversible by design.