This paper investigates how attackers can adjust the thresholds of their classification to optimize classification metrics in their membership inference attacks. By using Monte Carlo methods, we modeled the distribution of scoring functions for both True Positive and True Negative values. Then we calculated classification metrics (FPR, FNR, MA, and AR) as a function of threshold value and found a sigmoid relationship, verified by linearizing our data ($\mathbf{R}^{2}=0.997$). From this, we found relationships for most metrics as a function of threshold value and how to optimize them. We found that the relationship for FPR and FNR as a function of threshold value, T, follows a translated sigmoid function. Our findings provide information on methods attackers can use to fine-tune their thresholds to optimize their attack with minimal computational power. Our findings demonstrate the importance of altering aggregate statistics with Differential Privacy to mitigate Membership Inference Attacks. The major limitation of our model is that the attacker needs to know the underlying distribution of data, which we have assumed is Gaussian. In addition, we have only taken the case where data is binary. Additional research is needed to adjust for or reject these limitations.