We study pricing efficiency in decentralized prediction markets by comparing market-implied probabilities from Polymarket with benchmarks derived from option-implied risk-neutral distributions extracted from the derivatives market. We study Bitcoin prediction bets and find that, although Polymarket prices broadly track option-implied benchmarks, they show systematic mispricing driven by complexity, behavioral factors, and market frictions. Mispricing is most pronounced in tail events, during periods of high volatility, major macroeconomic shocks, and reflects the influence of sentiment, attention, and blockchain-specific risks. These results reveal both efficiency and behavioral distortions in prediction markets.