Ask HN: How effective is current driver assistance at avoiding accidents?

Looking at currently deployed driver assist technologies (things like emergency breaking, lane departure warning/assist/centering, speed limit recognition / cruise assist), assuming they become universal at some point where nearly every car has them -- what percentage of avoidable accidents is likely to be eliminated?

I'm thinking of things like a distracted driver doesn't see that traffic ahead has stopped, or the driver turns away from looking straight ahead due to a sudden noise and the car drifts into another lane. Or one step further, red light recognition that would keep a driver from accidentally running a red light. Would these be likely to only have marginal utility on the overall range of car crashes, or are they really game changers? Also assume various actions, such as audible alert only (red light, lane departure) vs automatic action (slamming on breaks when a car in front stops, or a kid runs out in the lane).

Also, if things like lane centering and traffic/speed limit aware cruise is active, how will this impact the ability for law enforcement to recognize and pull over an impaired driver?