Every attack has an expected value for the number of dice the attacker or defender expects the other to lose. For instance, in a 5 v 3, the attacker has a 90.9% chance of eliminating 3 dice for the defender, and a 9.1% chance of losing 4 dice for himself. The expected value from the attacker's POV is thus: (.909 * 3) - (.091 * 4) = 2.36 dice. The defender's expected value is -2.36.

If it were possible in each game to keep track the amount by which every player exceeded (or fell short) of the expected value of the rolls he participated in, you'd have a "gross luck" for that player. Dividing that value by the absolute value of the sum of all the expected values of that player's rolls would give you a luck "rate".

The theory behind using a rate is that having, say, five dice worth of bad gross luck in your first few rolls is probably enough to knock you from the game, while having that same amount of bad luck over ten rounds is hardly bad luck at all.

I'm inspired to suggest this idea by frustration over having obviously terrible luck in a game and not even getting the consolation of having the "luck" statistic confirm it.