2000 rolls is not enough of a sample size.
Well, it then be good for op to increase the sample size and probably reach a definite conclusion on whether your aligation is true or not. O P can equally send the me the data so that I can help him run a regression analysis and see what will be the out come of the result.
We need more people to willingly lend their data for further evaluation of what has been presented by the OP. But for now, the reputation of FortuneJack is still the same for me. They wouldn't risk their name with such "manipulation" in their system.