Why Full Information Beats Ranks

Two candidates have qualities drawn iid from a KNOWN Uniform(0,1) distribution, revealed as actual numerical values one at a time; you must accept one (reject-the-first forces the second). You win only if you accept the candidate with the higher quality. Compare the best win probability achievable by a full-information policy that uses the numerical value of candidate 1 against the best win probability of a rank-only policy that sees only relative order. Which is larger and by how much?