Should I expect to get a better answer from stepwise regression model if input variables are 20 with 550 observations and one output variable

I am using stepwise regression to generate a model. There are 20 input variables and 1 output variable with 550 observations. I am not getting good results. How should I proceed. If I get a p value greater than 0.05, should I remove it and again run the code or what should I do to get a better model. Is there any other criteria to remove the variable other than p value. I am a bit concerned about the durability of this stepwise function. Please enlighten me. Hoping for a reply.