How to asses expected value of X1>X2 after many permutations of not unique values?

I have n observed values. I permute them many times. Each time I only look at the values that show up on positions 1 and 2. Is there an expression that would give me expected value of P(X1%26gt;X2)? Known information about the values: number of all values, number of distinct values, number of repetitions for each value.How to asses expected value of X1%26gt;X2 after many permutations of not unique values?
Yeah?