Resolution Is Not Settlement, Part II: Protocol Finality and Observed Redemption on Polymarket

Battle Count: 50

Highly relevant to prediction market trading, event-linked derivative design, and settlement risk management. The paper provides empirical protocol-finality and redemption timing distributions (median 182-200 seconds) that directly inform capital lock-up costs, margin requirements, and conversion timing for leveraged event positions. However, it is primarily an identification and measurement study rather than a trading strategy paper. Its direct trading relevance is in understanding the finality gap between oracle resolution and cash realization, which affects position valuation, funding costs, and risk transfer in event-linked instruments. The state-occupation and capital-time estimands are inputs to margin and funding rule design (Paper 6).

Extremely high complexity: requires pinned ABI decoding, full blockchain event log processing over fixed block intervals, recursive block-splitting for provider-capped responses, exact canonical event identity management (chain/contract/tx hash/log index), deterministic question-to-condition mapping without fuzzy matching, lossless integer payout-vector decoding, right-censored survival analysis with multistate transitions, seven cumulative validity gates, zero-gap range ledgers, source-to-canonical deduplication, and network-disabled deterministic rebuild verification. The formal identification theory (13 proofs) adds theoretical complexity. Practical implementation requires deep blockchain infrastructure knowledge, event-sourcing architecture, and survival analysis expertise.

Exceptional reproducibility design: deterministic network-disabled rebuild from frozen inputs, content-addressed inventory checks, semantic verification of event identity, zero-gap range ledgers, ABI-derived event classification, full provenance lineage (source observation → canonical event → exact relation → analytical table/figure), version-locked hypotheses, negative controls, and source-generated result layers with input hashes. All analytical artifacts are derived from frozen datasets rather than manually entered values.