I'm convinced it makes a big difference, and least on arrows that tend to be inconsistant in specs (most qualify here). The only good way I know to do it, is to use a spine tester (fairly cheap to make one). You could always resort to turning nocks to get best groups, and cull arrows that you can't bring into the group. This would be a time consuming task requiring accurate shooting, but should end up working quite well.

I can't comment on the floating-in-water method. I've never tried it. It doesn't sound like it would be very accurate.