Is it logical to stand on the chance-line (50-50 %) when we don't know a-priori probabilities?

Let we have two hypotheses $H_0$ and $H_1$ and we don't know their a-priori probabilities.

If we wish to calculate the average probability of error, does it makes sense to assume 50-50 % chance of occurrence of any of two, in the context of Neyman-Pearson criteria?

• How is this related with roc? – Calimo Aug 28 '14 at 4:49
• I used ROC, just to take the viewer's thought towards thinking in terms of regions when we plot a ROC curve. For instance, line $P_d=P_f$ is called chance line, region below chance-line is termed as "biased-region" and region above chance line is where we perform intelligent-test, and ROC curve lies in this region. – kaka Aug 28 '14 at 13:23