Hi.

I would just suggest that you use the following algo for the second winner:
Remove the first winner from the list, in your example, this would result in the following list:

0 Jeremy

1 justinlarsen

2 loren

3 mail2mm

4 matthew_k

5 rearviewmirror (original 5 -oktane- was removed)

6 Thiemo

7 thrasher

And use a modulo 8 calculation. You algo would prefer the one following to the first winner, doubling his chances (because he would win on either 5 or 6 on your example), while mine would keep equal chances among the first round loosers.
Just wanted to add a little stochastic knowledge in here.

cu,
sven