In plain terms
Source Bias audit. Polymarket is the engine's primary prediction market data source.
Source bias audit. Polymarket is the engine's primary prediction market data source. Core biases: (1) Whale-dominated — single accounts can move markets by millions ($50 million+ individual positions documented). (2) US-restricted — American users technically barred, creating a non-representative participant pool skewed toward crypto-native, non-US speculators. (3) Crypto-denominated — all positions settled in USDC, meaning participants must already be inside the crypto architecture the engine tracks as a capture layer. (4) Liquidity-as-signal — thin markets can produce extreme probability swings that look like information but are just flow. The engine supplements with Manifold (open-source, play money + sweepstakes, retail crowd) but Manifold's play-money dynamic distorts incentives — participants don't lose real capital on wrong bets. Current status (Mar 30): Polymarket showing US-Iran ceasefire by April 7 at 11.5%, OpenAI federal backstop at 27%. These numbers are treated as signal but may reflect crypto-native sentiment more than geopolitical analysis. The engine should weight prediction market data as crowd sentiment, not ground truth.
Apr 2 live feed: Polymarket: 'Military action against Iran ends by May 31' at 84% YES — crowd pricing quick end. But Trump just committed to 2-3 more weeks AND threatened total grid destruction if April 6 deadline missed. The crowd is pricing resolution while the operational signals point to escalation. Hormuz monitoring protocol (Iran/Oman) gave markets brief hope — Dow recovered 600 points from lows. Prediction markets as sentiment, not intelligence, confirmed again.