How to Evaluate the Relative Likelihoods of Different Theories Being True if They Are Currently Untested

How do we evaluate the relative likelihoods of different theories being true if they are currently untested? Some physicists tend to use the word beauty to describe the characteristics of those theories that they think are most likely to be true. So it's not a question about what is beautiful, but does this sense of beauty as actually used in practice correlate with theories that are true? We're testing our theory about... Our theory about theories, we're finding out if our theory about truth is true, right?