Data sources and timestamps

Market data comes from a pluggable provider. Until a live provider is connected, a deterministic simulated provider generates a seeded random walk at six-hour steps; the footer states when data is simulated. In production the provider is CoinGecko (via MARKET_DATA_API_KEY). Every snapshot stores its source, takenAt, constituent version and formula version. Simulated values are never presented as live, and live values are never fabricated: if data is missing the narrative is marked unavailable.

Constituents and weighting

Each narrative has a versioned constituent set of five assets. Weights are equal and capped at 35% per asset; any excess is redistributed so a single large-cap token cannot dominate. Constituent sets are immutable during a live Race and every change creates a new version.

Narrative Score

narrative_score = price × 0.4 + breadth × 0.25 + volume × 0.2 + momentum × 0.15

• Price performance (40%): capped-weight average % change of constituents from the Race start, mapped linearly from -25% → 0 to +25% → 100.
• Market breadth (25%): share of constituents with a positive change × 100.
• Volume change (20%): average volume in the window vs the equally long prior window, mapped from -50% → 0 to +100% → 100.
• Momentum consistency (15%): share of snapshot intervals in which the weighted narrative index rose × 100.

All components are clamped to 0–100 before weighting. The pre-lock snapshot uses the seven days before the lock; Race snapshots measure from the Race start.

Missing-data rules

A narrative needs usable data for at least 60% of its constituents (minimum two). Otherwise it is marked unavailable, ranked last, and shown as “n/a”. If fewer than 60% of narratives have usable final data the Race cannot settle and must be voided by an admin.

Race Score

At settlement narratives are ordered by final Narrative Score. Starting rank is the pre-lock rank.

multiplier  = 0.5 + energy / 100          (0 → ×0.5, 100 → ×1.5)
role_points = min(cap, base × multiplier)
race_score  = leader + challenger + wildcard   (max 360)

Energy is virtual. Scores are produced by a pure function (scoreLineup) covered by unit tests. Settlement is idempotent: re-running it produces the same results and never awards XP twice.

Void Races

An admin may void a Race (for example, a data outage). Void entries award no XP, do not count toward accuracy, and neither extend nor break streaks.

Crowd aggregation

Crowd Picks aggregate human lineups only. AI lineups are excluded. We publish the most common pick per role, Energy-weighted conviction per narrative, a consensus lineup (top pick per role, de-duplicated, Energy re-scaled to 100) and the biggest disagreement (the role whose top two choices are closest). Crowd data is hidden from eligible players until their own lineup is locked or the Race has closed.

AI coaches

The three coaches are deterministic rule sets (strategy version ai-rules-v1), not machine learning. They draft from the pre-lock snapshot, lock at the deadline with the same three-role / 100-Energy structure, and store the snapshot id they used. They are never generated or revised after the Race starts.

• ROTATOR — ROTATOR ranks narratives by momentum consistency (60%) and volume change (40%) from the pre-lock snapshot, drafts the top two as Leader and Challenger, and picks its Wildcard from the highest-momentum narrative outside the top four.
• ATLAS — ATLAS blends normalized price performance (50%) with market breadth (50%). It drafts the top two as Leader and Challenger and picks the widest-breadth narrative ranked 5th or lower as its Wildcard.
• NOVA — NOVA compares the pre-lock snapshot with the previous week's snapshot. It drafts the two most-improved narratives as Leader and Challenger and picks the lowest-ranked narrative with a positive improvement as its Wildcard.

XP, levels and streaks

Levels: Observer 0 · Scout 250 · Analyst 750 · Strategist 1500 · Meta Hunter 3000 · Navigator 6000.

Badges

• First Lineup — Locked your first Race lineup.
• Winner Called — Your Leader finished 1st.
• Wildcard Master — A Wildcard gained four or more positions.
• Perfect Podium — Leader 1st, Challenger top 3 and Wildcard up in a single Race.
• Beat the AI — Out-scored all three AI coaches in a settled Race.
• Crowd Breaker — Beat the crowd consensus lineup with a different Leader.
• AI Specialist — Three correct calls involving the Artificial Intelligence narrative.
• RWA Specialist — Three correct calls involving the Real World Assets narrative.
• Gaming Specialist — Three correct calls involving the Gaming narrative.
• Three-Race Streak — Locked lineups in three consecutive Races.

Meta Rating and Meta DNA

meta_rating = round(avg_race_score × 4 + leader_accuracy × 300 + challenger_accuracy × 150 + wildcard_accuracy × 150)

Ratings are provisional until a player has three settled Races. Meta DNA needs the same sample and summarises best narrative, best role, early-discovery performance (Wildcard positions gained), conviction calibration (Energy placed on hits vs misses), consensus-vs-contrarian tendency (how often your Leader matched the crowd) and average finishing score. Labels: Early Hunter, Rotation Reader, Wildcard Scout, Consensus Navigator, Contrarian, Balanced Strategist.

Limitations

• Simulated data demonstrates the loop; it says nothing about real markets.
• Narrative membership is editorial. Reasonable people disagree about which assets belong where.
• Six-hour snapshots miss intra-interval moves; volume data from public APIs can be revised.
• MEGASPRINT is educational. It does not execute trades, hold funds, or give financial advice.