This paper presents an easier and new robust method for hypothesis testing to conclude significant mean differences between two independent or paired samples using the concepts of location, variability, confidence intervals and Gaussian distribution. For hypothesis testing of two samples, t-test is widely used. Beside this, Wilcoxon signed-rank test and often permutation test is also conducted. Each of these methods have their own rigorousness and drawbacks for which general people and non-statistics students often find it hard to conduct experiments using these. To fix these issues, a new method of hypothesis testing is proposed in this paper that basically utilises the properties of normally distributed data and resampling, and is relatively easier to calculate using only pen and paper. The time complexity analysis of each program is also conducted to give a concise overview about which hypothesis testing algorithm is more efficient and faster to execute, since statisticians use a lot of software nowadays for their analytical tasks.