If you really like the R^2 and its interpretation, you may want to consider the adjusted R^2 statistic that essentially penalized models that have more regressors. See Ezekiel, M. Methods of Correlation Analysis (1930) or almost any regression book. For the model y=beta0+beta1+...+betak, R^2_a=[(n-1)R^2-k]/(n-k-1).