I have a dataset of a list of 100. It is a 30-day retrospective average. Let's say I am confident that each value in the set has and R of .5 based on sample size. I can assign a historical value or score to each rank 1-100 based on actual, historical results. Now let's say I want to use this dataset as a forward looking tool rather than just calculation past results. If I know the R is .5, is there a fractional multiple I can apply to the scores that is reliable for a forward looking tool? For example, let's say the #1 rank in any 30-day average historically produces 3 points or whatever I want to call it. But I know the R value is only .5 so I'm therefore not 100% confident in the score and know there is some variance in it. Is there a simple way to account for this? For example, weight the score by 50%, 75% etc? So, if the score is 3 but i know the R is .5 I can take 3 * .5 =1.5 and use that confidently in my forward-looking tool?

I can regress the scores to the mean but that will not change the rank. Since the rank will determine the score, the score will not change after the regression.