If luck was not normalized to number of rolls, the statistic would suck even more. We know that a bad roll at the start might hurt most (I pledge to make a law against losing 4v2 attacks in 1st turn) - but then, "pain" or "significance to the game" are semantic values - and thus not computable.

But why are only attacks accounted for? In LUCKFEAST (i. e. 8v8 end game) winning 50% of your attacks (which would be more than you deserve) would make you lucky, even if you would defend not a single time. And honestly, nobody would consider this lucky. Thus defends should be in the statistics as well as they are as important.