Hi David, I believe the multinomial formula does not apply here, as we deal with the probability of a sum given n trials (over the combinations of outcomes that can generate the sum over n trial).
Given: number of trials = t, sum of random variables = Z, the discrete convolution is: