[Russell Sobie]

I get your point but, what you mention is the effect as it is now, my propose is directed to be a deterrent in order to be more caution in the first seconds after race start, that is the moment with highest risk because cars are all packed together.
I know that in case of incidents, that will happen as well, some driver would get an heavier penalization despite themselves, but the perspective of a rating and point loss will definitely reduce the probability of occurrence (can't say how much though), thus reducing the overall risk, with benefit for every driver.

So the net effect may be that SRS will start losing nice drivers while the bad ones will stay. And things will become worse.

Yup. The people you are trying to deter are not reading this forum, will not have read the rules, and will not pay attention to even an email sent out announcing that all lap one collisions will count 8x incidents or whatever. They will, as Pawel says, just race however they want and be quite happy that they finish in the top 10... even if their points put them at dead last.

I mean, personally I give zero shits about championship points. Or really even my driver rating when it comes right down to it. I race clean, give spots back on the occasion I screw someone over on accident, and get my kicks passing and getting passed in 20 minute races. Numbers on a webpage do not make me behave this way. My mentality does. The drivers you are worried about are going to care about these numbers even less than I do.

What if SRS could differentiate between "the front car" and "the back car" in rear-end collisions? Wouldn't that be better? Is that possible?

It would mean that, just like in real traffic (at least over here), the rear driver would be penalised and the front driver would go free... The idea is that anyone should always keep eyes front and be ready to brake...

I think by now this would have been done by some enterprising App creator for AC. It is likely any such system would have impossible-to-solve-for edge cases and false positives for poor behavior. I just recently had a collision with someone that was off track and drove directly into the front of me in a way that was impossible to avoid. To make it even harder to deal with systematically, it was a head on (he was facing backwards on the track at the time). Who was in front? Do you just look at who is "further along the track"? Do you look at where someone got hit? Too many variables, combined with all of the wacky stuff that goes on when someone has a really bad connection and is blinking in and out of other cars.

[Albert Castro]

It would be nice to have events that filter out users/racers based on incident averages. This is a bit of a repost from a thread that I started, but I took all the driver data on February 16, 2018 and just did a quick plot of "Driver Rating" vs. "Incident Average".
The results are not surprising. See below. total drivers/data points represented in the plot - top 5000 drivers, rating of greater than 80.