But you have another problem. You haven't really defined what a blessing is. This makes the experiment worthless even if you do the control.
What you need is something specific, something that an outside observer will agree is a "blessing". For example, if we had identical twins, inject both with a disease of our choosing (nonfatal of course), we could have one tithe and the other not. If tithing works, tithing twin should heal faster than nontithing twin. Then we have something.
But just having a vague "something good will happen" means nothing. It's too subjective. I like sushi, my brother hates it. So if we both get sushi, I record it as positive, my brother records it as negative. That won't work in an experiment. You need something objective and publicly observable.