Option pricing models are tools for pricing and hedging derivatives. Good models are complex and the econometrician faces many design decisions when bringing them to the data. I show that strategically constructed low-dimensional filter designs match and often outperform those that try to use all the available option data, in terms of state recovery, pricing, and hedging. The filters are built around option portfolios that aggregate option data, and track changes in risk-neutral variance and skewness. They also explicitly account for difficulties in the recovery of risk-neutral moments from option prices. The performance advantage is greatest in empirically relevant settings: in models with strongly skewed jump components that are not driven by Brownian volatility.