Safety Score

[FalconFour]

An appropriate course of action in this scenario, even with no Safety Score, would be to fall back as soon as you see someone signal to get in, to open up as much space as possible.

My problem is that there doesn't seem to be any sensible formula to it. AP is incredibly erratic about its following-distance, and once AP finally stabilizes when it's behind someone going slower than me with cruise control set, its distance at "7" is only almost imperceptibly different from its distance at "2" for example. Its following distance that it chooses at "7" seems too close to be "safe", and yet AP gets way WAY WAY closer (mentally FCW-inducing) when traffic is stopped ahead for example. So, given the inconsistency of how AP behaves, I can't find any way to be controlling the car above 50mph and trust that I won't get an "unsafe following" glitch for it. If it's dictated by AP's incredibly laggy logic, it could be virtually anything, and "the only winning move is not to play". In effect, the only "safe" following distance is infinite - the only way to be on the freeway is to use AP from onramp to offramp. So that's what I do. Disengage and die because AP is already following "too close"!

"Unsafe following is the proportion of time where your vehicle’s headway is less than 1.0 seconds relative to the time that your vehicle’s headway is less than 3.0 seconds."

You're quoting here the exact statement that I said makes no sense. Less than, and less than? It's like saying "Save up to 50% or more" - it's an infinite, unconstrained statement with no boundaries. What's it saying? Describe it in different words, maybe, if you understand it? When I read that statement, it read like a documentation error.

Replying here to your responses in the OFFICIAL BUTTON THREAD here, since it makes more sense to discuss Safety Score topics in the Safety Score thread.

Didn't even know this thread existed I only learned about it by this quote notification. Maybe ought to post it over in the Button thread so we can tie a bow on that one and throw the whole thing in the ocean.

 

[T-Racer]

I’m looking forward to seeing the Teslas in my neighborhood start to employ these advanced techniques. I’ve already seen several driving slowly around the neighborhood at night. As I was crossing the street on my mountain bike yesterday, I saw a Tesla at the top of a nearby hill apprehensively roll very slowly to a stop due to my presence. I know what they are up to. It’s probably great fun to roll out on a bike into the street in front of one - you just don’t want to cut it too close. Have to allow for slowing at less than 0.3g to ensure safety.

What you're describing sounds like Tesla drivers turning into Prius drivers. Then it occurred to me... maybe Prius drivers drive the way they do because they're all testing FSD-like beta software and sworn under NDA not to talk about it. Genius!

 

[HighZ]

a score designed for safety/collision avoidance ISN'T necessarily the best metric for FSD testing qualifications. Attentiveness could be.

I was sure that they would use the interior camera at least on the cars that have them to track some attentiveness measure. They just used a measure that was readily available and could be added to the app quickly. But I bet that the interior camera will have a role at some point as that is a good measure of attentiveness.

 

[HighZ]

So now we have to have a debate when they open it up to people by score, whether they are going to use the actual score, or the one with all this rounding error.

My guess is that they will take all the 100’s and sort them by total miles driven. Then 99’s, etc.

The other question is will they be looking at 7-day rolling averages or your cumulative score over the 30 days. The first may be easier if you have one really bad day you want to drop off.

 

[EVNow]

My guess is that they will take all the 100’s and sort them by total miles driven. Then 99’s, etc.

Within a given score - we have no idea how they will sort. Could easily be based on whoever pushed the button first. Could be a geographic / model distribution.

The other question is will they be looking at 7-day rolling averages or your cumulative score over the 30 days. The first may be easier if you have one really bad day you want to drop off.

My guess is either the current score at that point, or a snapshot of the score whenever they decide to pull a list. 7-day rolling is less likely (more difficult to program etc.).

 

[AlanSubie4Life]

Less than, and less than? It's like saying "Save up to 50% or more" - it's an infinite, unconstrained statement with no boundaries. What's it saying? Describe it in different words, maybe, if you understand it? When I read that statement, it read like a documentation error.

Maybe an example would help? Not sure. Let's say you follow behind someone for a total of 10 seconds. For the first 5 seconds, you are following at less than 3 seconds, but at greater than 1 second (this is following at less than 3 seconds). That adds 5 to the denominator, and 0 to the numerator. For the next 5 seconds, you are following at less than 1 second. This is BOTH following at less than 3 seconds, and following at less than 1 second. So you add 5 to the denominator, and 5 to the numerator. (0+5)/(5+5)

Finally at the end of the interval you change lanes so you are no longer following the vehicle, and the counter stops for both numerator and denominator.

This results in 5/10 = 50% for your unsafe following, as I understand it.

It's true that they don't explicitly say that headway cannot be negative, but that seems unnecessary given the context.

 

[dduffey]

You're quoting here the exact statement that I said makes no sense. Less than, and less than? It's like saying "Save up to 50% or more" - it's an infinite, unconstrained statement with no boundaries. What's it saying? Describe it in different words, maybe, if you understand it? When I read that statement, it read like a documentation error.

Following distance greater than 3 seconds doesn't impact the unsafe following score, so ignore that.

If you follow less than 3 seconds distance, that is your total "following time". If you then follow less than 1 second distance, that is the "unsafe" following time.

unsafe following score = time following less than 1s divided by time following less than 3s

So if you followed a car (less than 3s stopping distance) for 10s in total, but of those 10 seconds 6 seconds were within 1 second stopping distance then your score with be 6/10 = 60%

Following at greater than 3 seconds doesn't help because it doesn't count in the denominator.

If someone cuts you off closer than 1 second in distance and you immediately back off past 3 second follow distance you will have a high unsafe following score.

 

[HighZ]

If someone cuts you off closer than 1 second in distance and you immediately back off past 3 second follow distance you will have a high unsafe following score.

Exactly! This happened to me on a long drive when I was behind cars for a short distance and obviously too close. But the calc has a max of 60% and it is not as heavy of a penalty as others.

 

[AlanSubie4Life]

Following at greater than 3 seconds doesn't help because it doesn't count in the denominator.

If someone cuts you off closer than 1 second and you immediately back off past 3 second follow distance you will have a high unsafe following score.

Yes, one of the reasons I don't use AP on the freeway as much at the moment (less than I used to). I now camp out behind people (NOT on AP) at about 2.5 second following to rack up following time. I try to aim for at least 5 minutes, 300 seconds. (This is not improving safety - I like to not follow anyone whenever possible, or at least at 5-6 seconds or more if it is not crowded.)

That way if an unavoidable cut off occurs when I don't have time to engage AP (or on a fast surface street), I have a nice strong denominator and the result will be less than 0.5-1%.

Once I've racked up the time for a given day/trip, then I go back to using AP and don't worry about it as much.

 

[whazupman]

My problem is that there doesn't seem to be any sensible formula to it. AP is incredibly erratic about its following-distance, and once AP finally stabilizes when it's behind someone going slower than me with cruise control set, its distance at "7" is only almost imperceptibly different from its distance at "2" for example. Its following distance that it chooses at "7" seems too close to be "safe", and yet AP gets way WAY WAY closer (mentally FCW-inducing) when traffic is stopped ahead for example. So, given the inconsistency of how AP behaves, I can't find any way to be controlling the car above 50mph and trust that I won't get an "unsafe following" glitch for it. If it's dictated by AP's incredibly laggy logic, it could be virtually anything, and "the only winning move is not to play". In effect, the only "safe" following distance is infinite - the only way to be on the freeway is to use AP from onramp to offramp. So that's what I do. Disengage and die because AP is already following "too close"!


You're quoting here the exact statement that I said makes no sense. Less than, and less than? It's like saying "Save up to 50% or more" - it's an infinite, unconstrained statement with no boundaries. What's it saying? Describe it in different words, maybe, if you understand it? When I read that statement, it read like a documentation error.


Didn't even know this thread existed I only learned about it by this quote notification. Maybe ought to post it over in the Button thread so we can tie a bow on that one and throw the whole thing in the ocean.

I didn't get within 30 to 50 yards from another car today in ATL traffic. I think the car sometimes sees the car in an adjacent lane especially during a turn is too close. That's the only explanation how I got FCW dings today. I drove like a complete idiot from another planet and still got 4.2% FTC???

 

[nvx1977]

You're quoting here the exact statement that I said makes no sense. Less than, and less than? It's like saying "Save up to 50% or more" - it's an infinite, unconstrained statement with no boundaries. What's it saying? Describe it in different words, maybe, if you understand it? When I read that statement, it read like a documentation error.

"Unsafe following is the proportion of time where your vehicle’s headway is less than 1.0 seconds relative to the time that your vehicle’s headway is less than 3.0 seconds."

Yeah this confused me at first too. First, for this metric to count, you have to be following someone with 3 seconds of headway or less. If the car is further away from that, this metric is not even measured.

Once you are within 3 seconds of headway, the denominator starts counting time.

If during this period of time you get closer to the car where the headway is below 1 second, you generate the penalty numerator (also units of time).

So the final score is that fraction.

The more time you follow someone between 1 and 3 seconds, the larger your denominator and the less impact this penalty has.


-edit-
oops, prolly should have finished reading the thread before responding. sorry for the duplicate content!

 

[mhan00]

I had to go to Westwood today and did about 40 miles of freeway in LA traffic, one of my longer drives since the pandemic started and ended all my usual road trips. I was honestly a little concerned after reading the stories here, but I ended up with a 100 for the day. I don’t think I drove much differently than I did before (no way to say for sure, since even though I did try to drive normally I absolutely was thinking about the score in the back of my head while I was on the road). Now I’d be curious to sit in the cars of some people who are getting dinged to see how they’re driving.

 

[keith58]

I empathize with your plight, but don’t you think that Tesla will enter your Safety Score into evidence if you decide to go to arbitration and beyond?

I hope so. It speaks to Tesla selling a feature without a clear understanding of when it would be delivered, how it would function and with changing requirements and limitations on delivering what they contractually agreed to deliver. Better for me.

 

[dduffey]

I had to go to Westwood today and did about 40 miles of freeway in LA traffic, one of my longer drives since the pandemic started and ended all my usual road trips. I was honestly a little concerned after reading the stories here, but I ended up with a 100 for the day. I don’t think I drove much differently than I did before (no way to say for sure, since even though I did try to drive normally I absolutely was thinking about the score in the back of my head while I was on the road). Now I’d be curious to sit in the cars of some people who are getting dinged to see how they’re driving.

My Saturday score was 93. Today over 100 miles, 6 trips at 100. I am absolutely a worse driver today than in the first day (Saturday).

But now I know how it scores.

Haven't had to do a reset, but if I get cut off I now tailgate (1-3 seconds following distance). If I have an aggressive turn (had a turn on a highway exit and even then was honked at) I do some weaving to make up for it.

 

[kbM3]

I had to go to Westwood today and did about 40 miles of freeway in LA traffic, one of my longer drives since the pandemic started and ended all my usual road trips. I was honestly a little concerned after reading the stories here, but I ended up with a 100 for the day. I don’t think I drove much differently than I did before (no way to say for sure, since even though I did try to drive normally I absolutely was thinking about the score in the back of my head while I was on the road). Now I’d be curious to sit in the cars of some people who are getting dinged to see how they’re driving.

I got dinged twice the 1st day.

1. Short green light changed to yellow after only 2 cars went through and I was only at about 5-10 mph. It was also a turn, so if I ran it safely, I would have had to accelerate in the turn and harm my turning score. I chose to brake.

2. At 18 miles and a score of 92, I went for another 18 mile drive with lots of gentle deceleration to try and undo that bad braking score. But I had to park in a garage with a sharp ramp. I was taking it slowly, but a car entered coming the other way and triggered a Forward Collision Warning.

So rather than 18 miles at a 92, I made my situation worse with 36 miles at 90 for Day 1. FCW’s are DEADLY. i Have about another hundred miles since of perfect 100’s to get my score up to 98, but I’m now scared to death of front collision warnings in neighborhoods, parking lots…. Probably scarred for life

 

[Daniel in SD]

Now I’d be curious to sit in the cars of some people who are getting dinged to see how they’re driving.

I feel like I’m driving with them about half the time I’m using Autopilot in stop and go traffic. Then I change lanes in attempt to escape.

 

[FalconFour]

I just got a 1.0% on hard braking, among a sea of 0's on my last trip and every other trip today. ಠ_ಠ

I can't even. HOW.
So much for my one 100% day. Another day, another 99.

And that following distance thing, best described as "a window of 1 to 3 seconds of following distance". Knowing how incredibly laggy AP is to following a lead car, I don't believe I can trust those measurements. Still, the only safe way to play is not to play (use AP any time I'm going over 50). I'm still not confident in its ability to actually measure that properly, when it doesn't respond to slowing or stopped cars ahead, and even when I whip the speed setting down 5 times (was set:70, going:65, new target:45), it lazily slows down (64, 63, 62...) until it damn near rams into them and slams on the brakes. Really don't think I can trust Tesla's judgment of what following time is, especially with a seconds-based timer!

 

[TIppy]

I do some weaving to make up for it.

Seems perilous. How do you know you're pulling more than 0.2 and less than 0.4 Gs? Do you have a smartphone g-meter?

 

[queue]

Anyone figured out if we can go down a hill without triggering hard-braking ? With AP / no AP or low regen ... ?

Need to visit the low-lands - wondering whether to take the Tesla or the other car

don't brake and run all stop signs and hope there aren't any sharp curves.

 

[TIppy]

I now camp out behind people (NOT on AP) at about 2.5 second following to rack up following time.

Man, that seems like you're tempting fate for a FCW, which are painful.