I can be wrong, but as far as I can tell octane number is something that is determined experimentally, I am not aware of any simple rules that can be used. Could be they exist, but if so they probably can be used to compare two fuels with assumption that everything else is identical - say n-heptane and n-octane, or more branched octane with less branched octane. They probably fail miserably when comparing two completely different compounds (like say n-hexane and branched octane). That's just an (more or less) educated guess, but that's what I would expect.