For all three of my roulettes, I've done a stack-rank scoring method. I figure, it's like the 100m sprint. Doesn't matter if you ran a 9.7... when someone else runs a 9.68, you are in 2nd place. Once you get past the mind-set that being in 10th place doesn't mean that your song was 10x worse than the 1st place song, I find it much easier to score. You also don't have to think about scoring consistency across rounds... you just have to rate the songs best-to-least best.
But yes... in the end, my last roulette basically confirmed that the stack rank method ended up with virtually the same result as the #/10 method. I think two people switched spots, that was the only difference.
My suggestion Splent ... keep it simple. There's merit in the KISS principle, especially for someone's first roulette. Fire off some themes, rate the songs, and declare a winner after 8 or more rounds. Complexity is usually more fun for the organizer than for the players - not a shot at anyone in particular, other than myself for my "Jeopardy" style Roulette v2. Caught a lot of playful grief over that one.