Kernel techniques for efficient error probability estimation have demonstrated good performance in Bit Error Probability (BEP) estimation under a diversity of simulation frameworks. However, the transmission schemes of contemporary and emerging communication systems are more and more complex so that it is questionable if the Symbol Error Probability (SEP) estimate can be anyway derived from the BEP estimate. In this paper, we investigate for a direct way to efficiently estimate the SEP based on the kernel technique. We focus on the ubiquitous Quadrature Amplitude Modulation (QAM) systems and design a Gaussian kernel-based estimator. Simulation results, involving 4-QAM and 16-QAM symbols transmissions over the additive white Gaussian noise channel and over a frequency-selective channel, show that the proposed estimator can perform efficient, reliable and accurate SEP estimations.