Creating a Score to Rank Single Variable Regression Results

[hedgie]

I have run single variable least square regression across a number of different securities. I would now like to create a score to rank the regression results. I am using three outputs from the regression as my ranking criteria: 1) R Squared, 2) Beta, 3) P-Value of Beta. So subjectively, a security with the highest r square, highest beta, and lowest p-value will have the highest score and be ranked first.

I am currently using the formulae: (rsquared * beta )/ p-value. By dividing the product of rsquared * beta with the p-value, I am trying to adjust for the statistical significance of the P-Value. Any suggestions on the arithmetic for the aggregate score? Does this make sense?

The other way I can create a score is to rank each output and sum up the ranks, with the lowest being the highest ranked i.e. a security with the highest r square, highest beta, and lowest p-value will get rankings of 1. If this turns out to be the same security, then this security will get a score of 3, which will be the lowest in the sample set and rank first. Does this option make more sense to use?