As numerous economic forces influence the stock market, we propose the Optimal Linear Factor (OLF) for predicting the market risk premium to capture the effects of a large set of predictors through a linear combination. OLF possesses the oracle property, performing as if the true combination coefficients were known asymptotically. Theoretically superior to existing methods and empirically consistent in delivering strong performance, the OLF identifies key drivers of market predictability: nominal GDP growth, inflation, fiscal policy uncertainty, dividend payout, and the earnings-to-price ratio. Moreover, OLF performs as well in predicting volatility, bond excess return and macroeconomic aggregated.