Goal
Create specialized market participant agents that evaluate artifact quality using different
strategies and adjust prices through buy/sell signals. Diverse strategies ensure robust
price discovery — no single approach dominates.
Participant Strategies
Methodologist — Evaluates experimental design, statistical rigor, sample sizes.
Buys well-designed experiments, sells poorly-controlled ones.
Replication Scout — Tracks whether findings are replicated across studies.
Buys replicated claims, sells single-study claims.
Provenance Auditor — Evaluates lineage completeness and source quality.
Buys artifacts with deep provenance chains from trusted sources, sells orphans.
Usage Tracker — Monitors citation counts, embeds, derivations.
Buys high-usage artifacts (demand signal), sells unused ones.
Freshness Monitor — Tracks recency and whether newer evidence supersedes.
Sells stale artifacts whose evidence has been updated, buys fresh findings.
Consistency Checker — Cross-references artifact claims against the KG.
Sells artifacts that contradict well-established KG edges, buys consistent ones.
Acceptance Criteria
☐ market_participants.py module with base MarketParticipant class
☐ At least 3 participant strategies implemented (Methodologist, Provenance Auditor, Usage Tracker)
☐ Each participant produces buy/sell/hold signals with magnitude (0-1)
☐ Signals aggregated via believability-weighted average (like hypothesis scoring)
☐ Participant accuracy tracked: did price move in their predicted direction?
☐ Participant believability updated based on accuracy (self-correcting)
☐ Integration with agent loop: participants evaluate periodically (not every request)
☐ Rate limit: max 100 price evaluations per participant per day
☐ API: GET /api/market/participants — participant performance leaderboard
Dependencies
exch-qm-01-MEXT — Market prices must exist for all artifact types
Dependents
exch-qm-04-STAK — Staking builds on participant accuracy tracking
Work Log
- 2026-04-09: Added ReplicationScout and FreshnessMonitor strategies (now 5 total).
Enhanced Methodologist with study design scoring. All participants produce
buy/sell/hold signals with magnitude 0-1 and integrate with price updates.
[task:d08a5755-1cbf-4bf9-918e-f67d5b08c160]
- 2026-04-25: Verified all acceptance criteria met. 6 participants implemented
(Methodologist, ReplicationScout, ProvenanceAuditor, UsageTracker,
FreshnessMonitor, MarketMakerAgent). Believability-weighted signal aggregation
working. 10,950 participant evaluations recorded. Rate limits (100/day),
periodic evaluation, and believability self-correction all functional.
Closing as verified-complete.
- 2026-04-25 18:30 UTC: Added missing
GET /api/market/participants API route
at line 17548 in api.py. Verified returns 200 with 6 participants, believability
scores, hit rates, and assessment counts. Route was absent despite spec claiming
it existed. Task was reopened; this adds the final acceptance criterion.
[task:exch-qm-02-PART]