# Can R-square be a criterion of simple linear models?

For example, I constructed three simple linear models, say

Y ~ A

Y ~ B

Y ~ C

A, B and C are highly correlated

Now I have their R-squares and P-values, can I say that one with highest R-square and lowest P-value is better than others? If can't, what criterion should I use? My background is ecology BTW, I seldom got R-squares higher than 0.3

$$R^2$$ is a function of mean squared error (MSE), a value that we hope is small.
$$R^2=\dfrac{ \sum_i\big( y_i-\hat y_i \big)^2 }{\sum_i\big( y_i-\bar y \big)^2}\\=\dfrac{ nMSE }{\sum_i\big( y_i-\bar y \big)^2}$$
Whichever model has the highest $$R^2$$ also has the lowest MSE. Consequently, there is a sense in which that model has the strongest performance, yes.
Because of this setup where you have the same $$y$$ and same sample size (really, it’s the latter than matters), the highest $$R^2$$ will correspond to the lowest p-value on the slope parameter, as the p-value is a function of the MSE and the sample size.