That doesn't mean it DOES penalize you for improperly turning AP on--- just that if you do so AND something that'd cause a negative score happens- the system will 'count' it, which it wouldn't do if it happened in "proper" AP use.
So are you saying you don’t think the scores are masked on streets with traffic lights, etc.?
It really is a mystery; the documentation leaves Tesla wiggle room, and there are reports in both directions about how this actually works (in reports there is a lack of clarity about whether the use of AP was proper). Of course no one on the internet can be trusted so the mystery persists.
I’m not sure how it’s not penalizing to count a vehicle-initiated action while on AP in the score, but I guess that’s semantics?
There are two choices:
1) Drive yourself on a street with traffic lights, smoothly, get a decent score.
2) Use AP “incorrectly” in this scenario, have it jerking the brakes in response a some stimulus which was easily anticipated if you had been driving manually. Get a demerit.
Seems like if 2 was done with the expectation that the incident would be masked, but Tesla has decided to not mask those scenarios, that would be “penalizing” for using AP in an unsupported scenario.
If those events aren’t being masked, I’m not sure the value of the AP miles accrued is worth it. They say they calculate a mileage-weighted safety score, and those miles would help, but depending on what that means, it could only help with the FCW events (which are rare). It’s not exactly clear what they mean by mileage-weighted safety score though. If they really are condensing down each score on a day and then weighting those scores over the days by the number of miles traveled on each day (how I would interpret that literally), then yes, additional miles on the freeway on AP with actual masking of all events, except forced disengagement, could be a real benefit in offsetting a bad day. That would give a very different result than looking at the overall braking event %, following distance %, etc, over a 30-day interval, and then calculating a score just one time from all of those component percentages and rates.