We present an embarrassingly simple method for supervised learning of SABR model’s European option price function based on lookup table or rote machine learning. Performance in time domain is comparable to generally used analytic approximations utilized in financial industry. However, unlike the approximation schemes based on asymptotic methods – universally deemed fastest – the methodology admits arbitrary calculation precision to the true pricing function without detrimental impact on time performance apart from memory access latency. The idea is plainly applicable to any function approximation or supervised learning domain with low dimension.