Poll: current safety score

[FalconFour]

This was my attitude with Smart Summon and NoA. Both perform poorly for me so I stopped using them. But it was cool to try it out when it first came out.

If I get into this FSD beta early access, I'll try to use it on a regular basis because I know it's likely my usage is helping development in some way. I don't think using smart summon or NoA is improving either feature at this point.

oh man. Smart Summon is kind of a joke, mostly because of how unreliable it is to activate - the app probably connects/works about 25% of the time, so little that I just anticipate it not working and don't even try to show it off anymore.

But NoA? Very useful. Always enabled. I've learned that some settings are just there for show (like no-confirm lane changes by vibrating the wheel - it actually makes it slower to confirm), and that it often makes bad decisions. But that's why I just have a habit of noticing the little blue "I want to change lanes soon" popup, and tap it away if it's not valid. It's excellent for being in the correct lane at the correct time, and great for unfamiliar areas. Just enable confirmations (by means of not enabling no-confirm changes--... wait, is this a triple or quadruple negative now?), and when you see it offer a valid change, tap the drive stick to accept it and do it. It mostly does a great job.

 

[FalconFour]

Folks worrying about the results of this poll being representative of their wait time... just chill and see. Good chance that us super-active Tesla stans (able to vote in this poll, with less than 250 responses today) are kinda in the minority, and there aren't tens of thousands of people gaming the scores. I expect a lot of fun to be had in the coming weeks.

 

[HighZ]

I got an 81 on autopilot today, which is 15 points lower than I ever got driving manually.

That is the problem with a lot of AP driving. Nothing on AP (supposedly) is used in the safety score other than forced disengagement.

So you are essentially compressing the safety factor scoring into only that time off of AP, which is probably when we all have a greater risk of following too close, turning quickly or braking hard. That last one is the worst by far.

 

[nvx1977]

oh man. Smart Summon is kind of a joke, mostly because of how unreliable it is to activate - the app probably connects/works about 25% of the time, so little that I just anticipate it not working and don't even try to show it off anymore.

But NoA? Very useful. Always enabled. I've learned that some settings are just there for show (like no-confirm lane changes by vibrating the wheel - it actually makes it slower to confirm), and that it often makes bad decisions. But that's why I just have a habit of noticing the little blue "I want to change lanes soon" popup, and tap it away if it's not valid. It's excellent for being in the correct lane at the correct time, and great for unfamiliar areas. Just enable confirmations (by means of not enabling no-confirm changes--... wait, is this a triple or quadruple negative now?), and when you see it offer a valid change, tap the drive stick to accept it and do it. It mostly does a great job.

I have heard many people rave about how good NoA is for them. It is almost polar opposite for me. I suspect it has to do with interchange designs and local driving aggressiveness differences. NoA frequently misses exits, particularly when multiple exits are close together, or lane merges are weird around them. It's just too unreliable. It is also too timid to merge into the right lane. Here, we have right lanes that back up for miles for major highway interchanges while the other travel lanes are clear. People are cutting in last minute to save time. NoA will turn on its blinker forever and never find a spot. Also, when I need to do a manual lane change because I can anticipate a problem before a car, there's a huge delay between the signal and the car moving over. That delay is significantly reduced on regular AP.

So essentially, the stuff that NoA offers above regular AP doesn't work well here. For me. So I just stick with AP. I had very high hopes for NoA. I have demoed it to friends considering EVs, and several times I got egg on my face when NoA flopped on an easy exit. Here's one. I have no idea why it bombs this exit, but it fails every time:

Google Maps

Find local businesses, view maps and get driving directions in Google Maps.

www.google.com

Either it never makes the turn to get into the exit lane, or it tries to run right into the guardrail at the last second.

 

[DaveG_NJ]

I have heard many people rave about how good NoA is for them. It is almost polar opposite for me.

Same here. It's very good when there is no traffic or only light traffic. In the NE however, traffic is always heavy and drivers are aggressive, haphazard and, for the FSD brain, unpredictable. The car's computer cannot take a long view of the road and has no degree of situational awareness other than what is immediately in front of it.

I have a similar exit ramp that backs up for over a mile every morning. You either know that and get in line, or you slide up in the adjacent lane, looking for a sucker who will leave a gap and you take that spot. NoA just throws in the towel.

 

[nvx1977]

It's very good when there is no traffic or only light traffic.

unfortunately I can't even say that is the case for me. I have been out at night when there are very few cars, and the car still misses exits. Notably 495NB to Rt 3 NB (would love to know if others in my area can corroborate), and the exit I posted above (93SB to 213 loop connector). Typically I will see on screen that it wants to take the exit, but when it gets there, it just drives right on by with the blinker on. If there happens to be an A/B type exit and these are close together, it sometimes misses A and takes B instead. I think it's just really poor interchange designs around here. Although this summer I spent 2 months in LA and saw different but equally stupid interchange designs as well. I didn't try NoA there, so I can't say if it does better. I can't imagine it does.

 

[MrYenko]

Same here. It's very good when there is no traffic or only light traffic. In the NE however, traffic is always heavy and drivers are aggressive, haphazard and, for the FSD brain, unpredictable. The car's computer cannot take a long view of the road and has no degree of situational awareness other than what is immediately in front of it.

I have a similar exit ramp that backs up for over a mile every morning. You either know that and get in line, or you slide up in the adjacent lane, looking for a sucker who will leave a gap and you take that spot. NoA just throws in the towel.

My commute has a similar issue where the exits back up the right two or three lanes essentially solid for six or eight miles before my exit. A human driver will stay left until clear of the exit before theirs. NoA will happily tuck into the right lane and crawl along at 7mph for miles.

 

[Hydrocarbon302]

We should do a poll on how many miles you logged the past week. 318 miles logged here, and an overall score of 99.

you‘d think between someone who drove 10 miles and parked their car for two weeks with a 100% score and someone who drove their car 2000 miles with a score of 97%. The person with the 97 percent score would get FSD beta first.

I’m only saying that because I’ve got a 98% after driving 1543 miles since September 25th… and it kinda makes sense too.

 

[jebinc]

you‘d think between someone who drove 10 miles and parked their car for two weeks with a 100% score and someone who drove their car 2000 miles with a score of 97%. The person with the 97 percent score would get FSD beta first.

I’m only saying that because I’ve got a 98% after driving 1543 miles since September 25th… and it kinda makes sense too.

It does, make sense; but reality may not.

 

[Steve125]

you‘d think between someone who drove 10 miles and parked their car for two weeks with a 100% score and someone who drove their car 2000 miles with a score of 97%. The person with the 97 percent score would get FSD beta first.

I’m only saying that because I’ve got a 98% after driving 1543 miles since September 25th… and it kinda makes sense too.

This seems to be more about marketing and casting the perception FSD is getting closer and worth buying. It's just YOU that's not quite ready yet now.

 

[Callahan27]

Mileage per day will have to play a part of the decision. I can’t see them letting someone into the Beta without much mileage to validate their safety score. We should know soon enough

 

[AlanSubie4Life]

As mentioned elsewhere, we really need to start tracking unrounded (well...rounded to first decimal place) scores if we want to understand current score and queue position. Wonder if there will be an app update this week for that? I would think whether Tesla does release an update will depend on whether the number of 100 scores exceeds 1000 substantially.

 

[EVNow]

As mentioned elsewhere, we really need to start tracking unrounded (well...rounded to first decimal place) scores if we want to understand current score and queue position. Wonder if there will be an app update this week for that? I would think whether Tesla does release an update will depend on whether the number of 100 scores exceeds 1000 substantially.

Irrespective of whether they update the app or not - they can always use unrounded numbers to calculate the queue position of everyone. "Order By SafetyScore Desc" would be so easy ... unless they are rounding and storing, instead of rounding when displaying

 

[AlanSubie4Life]

Irrespective of whether they update the app or not - they can always use unrounded numbers to calculate the queue position of everyone. "Order By SafetyScore Desc" would be so easy ... unless they are rounding and storing, instead of rounding when displaying

They keep all the data AFAIK so I don't see any issues with what is tracked by Tesla. I wasn't commenting on their ability to track queue position!
Of course they can do whatever method of ordering they want. But it's one thing to do that, and another thing to not provide guidance to people as to what exact rationale they used. (Which they have not, so far. And users are not really proficient at calculating their actual score. So for the most part people will likely get predictably upset for entirely predictable reasons.)

 

[LastGas]

I'm glad it's over. I got a miserable 92 almost all because of HARD braking, none of which was strong enough for me to actually feel. But I have the State Farm Drive Safe and Save app tied to an accelerometer. It shows me on the map exactly where the events were: slowing down when the speed limit changes, traffic signals and turns at intersections where the speed limit is 45/55. Even being intentionally gentle didn't fix them all.

It would have been fairer if we got a week to practice because this "safe driving" is not normal driving.

 

[Hitmanblood]

I'm glad it's over. I got a miserable 92 almost all because of HARD braking, none of which was strong enough for me to actually feel. But I have the State Farm Drive Safe and Save app tied to an accelerometer. It shows me on the map exactly where the events were: slowing down when the speed limit changes, traffic signals and turns at intersections where the speed limit is 45/55. Even being intentionally gentle didn't fix them all.

It would have been fairer if we got a week to practice because this "safe driving" is not normal driving.

Whats over? You have to keep going to October 8th now. It's 14 days now according to Elon. Gotta love driving like a grandma and running lights so you don't get dinged. This "safety score" is anything but safe.

 

[aronth5]

I just hope Tesla doesn't expand the FSD beta drivers by 1,000 new drivers each day. Prefer to see 1,000 per week and then after 2-3 weeks increase the rollout if FSD testing by new beta drivers is going well.

 

[kaffine]

I'm at 76 overall so far. If I stay off the freeways I normally get in the 90s but I always get dinged for aggressive turning.

There is also an intersection I go through often that it just freaks out. It normally freaks out about lane departure and collision warning. It is a 5 way intersection and there is a tall curb/center divider you drive at and the turn but the car thinks you are departing your lane and that you are about to hit a wall. I normally have TACC set when going through but never have lane assist on going through there.

I have stopped looking at the score much. I'm not going to change how I drive to try and get the score up. However I have though about putting the car up on my lift and setting cruise at 70mph and just letting it run for a few hours. Would really help my kwhr/mile rating as well. Although I am sure since I have a RWD it would freak out about the front wheels not spinning. I know summon stops quickly if it doesn't see all the wheels spinning during use.

 

[Guruguys]

I will say, just by way of commentary that:

1. I started at 100 after a 50 mile drive with mostly autopilot
2. dropped to around 92 after the 50 miles to return home and fighting Houston traffic, mostly with hand driving
3. clawed back to 97 with 50mi of granny driving, by hand in a suburban, surface street environ

It's really not tailored for Houston. With about 700 miles now I have 98 overall but I have to put on autopilot most of the time and just let it brake hard for me or I get penalized. Trying to turn at low speeds has Houston drivers on ass entire time - it's not 'natural' here, and 'following too close' seems really inconsistent from what autopilot does vs distance you drive manually and get penalized for (ie. Drop out of autopilot following distance and instantly get dinged for following to close).

 

[old pilot]

Whats over? You have to keep going to October 8th now. It's 14 days now according to Elon. Gotta love driving like a grandma and running lights so you don't get dinged. This "safety score" is anything but safe.

I don't think this is correct. On the 17 th Musk stated that the test would only last seven days. On the 27 th he stated an apology for implying that FSD 10.2 would be available at the end of the test period when it would not be available until the 8 th. He only changed the release time for the 10.2 update not the length of the test period. Am I mistaken?