-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Safety Score
- Thread starter ScottNY
- Start date
sperkin
Active Member
Safety score shouldn’t be calculated when on AP. Maybe it was while you were off autopilot.
Tam
Well-Known Member
I just test drive with 2 trips and each is 3 miles for a total of 6 miles all on Autopilot except for home garage/business parking lot/turns at intersections and I got 100/100 each of the time and also combined for the whole day so far.
Last edited:
My current safety score on my 3 is a button shattering 99. My poor wife‘s score on her S is a mediocre 98. It’s nice to remind her who is the good driver in our household as she has spent years criticizing my driving.
I hope the button is causing similar clarification in all of your homes. Men are naturally superior drivers and it’s about time technology proves it.
I hope the button is causing similar clarification in all of your homes. Men are naturally superior drivers and it’s about time technology proves it.
MarcG
Active Member
Stop and go is likely to be under 50mph. Per Tesla’s description of the unsafe following, it is only measured when driving above 50mph.
The other thing I noticed is that it’s measured as a percentage of time spent under 1 second behind another vehicle versus the time spend between 1 and 3 seconds behind (but not beyond 3 seconds).
So ideally you always want to follow another vehicle at a far enough distance that it’s over 1 second ahead, but not beyond 3 seconds ahead, while driving over 50mph.
This will ensure your denominator is very large, so thet even if you follow someone “unsafely” (within a seconds) for a bit, the percentage will be low.
Last edited:
EVNow
Well-Known Member
sperkin
Active Member
AlanSubie4Life
Efficiency Obsessed Member
I feel that there is some ambiguity in Tesla’s description that “Driving on Autopilot will not be included in Safety Score calculations.” They may be simply stating that using AP just is not one of the factors (like % of time on AP or whatever) - not that things like unsafe following are not included in the score if AP/TACC is engaged.
Some here have claimed that using minimum follow distance in AP resulted in demerits. I have not had a chance to check myself nor do I want to!
Should not be hard for someone with a low score to do a definitive experiment above 50mph with follow distance 1 though.
It seems like the hard braking criteria encourages rolling through stop signs and not slowing as much for speed bumps. I wonder if FSD beta is evaluated on similar metrics?
Also unclear if braking to disengage Autopilot to get out of a situation it didn't predict then results in a lower score as technically Autopilot is no longer engaged?
Also unclear if braking to disengage Autopilot to get out of a situation it didn't predict then results in a lower score as technically Autopilot is no longer engaged?
AlanSubie4Life
Efficiency Obsessed Member
Thanks, missed that (although I think I saw someone post it earlier). That seems clear enough. I guess people will claim what they will…
Of course it is also possible that the close following observed is an error on Tesla’s part. 23.1% is a bit higher than I would expect for a drive mostly on AP with following distance 7 (what I was responding to).
I guess since the person was in stop-and-go traffic that would explain it. Denominator very small as mentioned above (since AP time is supposedly ignored, and low speed would be ignored). So close following occurred at other times probably.
Last edited:
Daniel in SD
(supervised)
The question is whether the ratio of time spent following <1s to <3s is an "event"?
There's also the question of what "appropriate utilized" is? Maybe "events" that occur off of divided highways count?
It’s not showing for me, either. I update the app and logged out and back in with no success.
I got a “hard braking” on my first trip. I’m pretty sure it was AP. It stopped short at a light, to the point where my foot was over the brake and I was shaking my head in disappointment (I rarely use it because it has poor driving habits). It was easily the hardest brake of my few mile drive.
Does anyone know of an app (iPhone) that can record g-force and play a notification when over a user set limit?
Does anyone know of an app (iPhone) that can record g-force and play a notification when over a user set limit?
I had to make a second drive. I reset reset the car before the second drive, and when I got back ensured it was really on wifi before I got out of the car.
AlanSubie4Life
Efficiency Obsessed Member
Seems like more experimentation is needed after discussion above. Seems like setting follow distance to 1 and using AP a bit for a short segment over 50mph would be interesting.
I’m not entirely convinced Tesla is calculating the overall averages correctly either. They can’t average the percentages of course and they may be ignoring 0% in the individual segment reports.
For example, I have taken 5 segments today. The first and the third were the longest, equal length, 0% score on one, 7.3% on the other long one. The fourth segment was fairly short and I got a 2.4%. My average is suspiciously 4.9%. This is the average of 2.4% and 7.3%. (4.85%, an invalid calculation in any case most likely; you cannot average averages)
Mathematically it is possible that this is a coincidence due to the way events and the ratios are counted…but suspicious:
(All approximate below, does not quite add to 46 since they do not report individual segment lengths I had to estimate.)
Trip 1: 11 miles. 0%. Freeway, lots of following events.
Trip 2: 2.5 miles. 0%, no events, surface streets.
Trip 3: 13.5 miles. 7.3%. Freeway, lots of following events.
Trip 4: 7.7 miles. 2.4% freeway, small amount of following events.
Trip 5: 10.1 miles. 0%, freeway, lots of following events.
Something to keep an eye on anyway. It is a Beta after all. With three numbers rather than two non-zero if it is still the simple average of the averages that would be increasingly unlikely to be correct unless all the segments are very similar… (others with multiple non zero segments could also check the averages…)
Last edited:
Similar threads
