It has always been the concern of side channel analysis that how to recover the key with a probability of about 1.00 under the condition that the number of power traces is very small and the success rates is very low. In order to recover the key, the attacker has to try to reduce the guessing entropy to decrease the uncertainty of the key. Unfortunately, guessing entropy is only a evaluation of attack ability in most cases. In this paper, we introduce the statistical characteristics of guessing entropy and propose guessing entropy enhanced CPA (GE-CPA). Its feasibility is verified in theory and experiment. Experiments on both AES algorithm implemented on an AT89S52 single chip and power trace set secmatv1 of DES encryption on the side channel attack standard evaluation board(SASEBO) from the website DPA contest v1. The experimental results show that, by only repeating the experiments less than 30 times, our GE-CPA can effectively recover the key even under the bad condition that success rate only ranges from $$5\,\%$$ to $$8\,\%$$. Thus, the problem is well solved.