It could mean that more people are opting in... Like the smart people wait to click the button until tomorrow, and do a single 0.1 mile 100% drive.
-
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
AlanSubie4Life
Efficiency Obsessed Member
We're never going to ascertain that in this thread. However, I'm not really making a giant mental leap here though - there's very little doubt in my mind that people who operate closer to the limit of traction for their tires are at higher risk of an accident, when looking at an enormous dataset.
I'm pretty sure other insurance companies measure this the same way and I'd imagine they have some data to support these metrics, just as Tesla does.
A driver who deliberately allows extra distance to stop on a downhill because they know their tires are working harder, or don't like being pulled forward in their seat excessively, is probably going to be safer, on average. Or low measured g forces may simply indicate that they don't live in a hilly area (which would be quite hazardous in winter). That would be fine too. Mission accomplished.
I don't want to do that. As mentioned elsewhere, I feel like people should just get maybe one mulligan, maybe as many as 3, before resetting the eval period to another two weeks (two weeks seems appropriate
). Made it safely to work, 0.0%/0 everywhere.
So just got the app on Android. Not a whole lot of driving this week so far. Mostly my wifes driving. I think I have driven the car about 20 miles this week. About 150 miles total and we have a 99 score. Looks like only dings on Unsafe Following. Tomorrow we go on a road trip and I will do most of the driving. See how it changes with my driving.
Oh also does it does it not take into account Autopilot. I tend to use Autopilot far more then my wife. So tomorrow will be mostly on Navigate on Autopilot.
Oh also does it does it not take into account Autopilot. I tend to use Autopilot far more then my wife. So tomorrow will be mostly on Navigate on Autopilot.
EVNow
Well-Known Member
Not the point. By being more selective - and adding more parameters you get better models.
You may see that people who brake harder on flat ground are more likely to get into accidents than those who brake harder only on downhill. Don't you think this is a reasonable assumption to be tested ?
ps : Doesn't Florida have more accidents than Seattle
But it's also not a leap to recognize there is a curve here. The person that drives 15 MPH under the speed limit and crawls around every corner is not the safest driver, despite never using 0.1G of their tires. There is a middle bound, and all drivers at the extremes are likely less safe.
The fact that an algorithm like this treats each input independently and without context makes it pointless.
You'd hope, but that doesn't make it true.
Just a question- how do you have a picture of a WRX cornering hard as your profile but believe a single measure of a single hard brake event is a reasonable measure that you are a less safe driver than someone else?
Context needed. Plenty of places have lots of hills yet very temperate climates with zero snow.
AlanSubie4Life
Efficiency Obsessed Member
I was never saying that a better model that more comprehensively measures safety was not possible. I was very narrowly stating that it's correct that the hard braking score measures how hard the tires are working; it doesn't need "cleansing." If you want to take the braking data from the vehicle and break it down by WHY the vehicle was braking hard, and draw conclusions about relative safety in different driving environments - that can be done with the dataset, and there's nothing wrong with that. It might be pretty interesting.Not the point. By being more selective - and adding more parameters you get better models.
You may see that people who brake harder on flat ground are more likely to get into accidents than those who brake harder only on downhill. Don't you think this is a reasonable assumption to be tested ?
ps : Doesn't Florida have more accidents than Seattle
Environment is another variable. I'm talking about comparing measured braking forces amongst people in identical environments.
If I'm gonna give Elon credit for something else, it's overhyping. I'm guessing he/they knew the rollout strategy beforehand and then Elon said a bunch of things in public to increase everyone's excitement. Things that are 10x better than what he knows the reality will be.
Either he's consistently wrong or he's consistently deliberate. I'm gonna go with deliberate; works every time, and 'reasons' can always be made up to explain the setbacks...
EVNow
Well-Known Member
Thanks for all the speculation - written as fact.
I'll take you more seriously if you can list all the people who don't "overhype" but built companies like Tesla & SpaceX. It takes a certain unrealistic / overly optimistic personality to put all the money he got selling Paypal into industries that had zero history of success.
AlanSubie4Life
Efficiency Obsessed Member
Yes, I admit it, that's me. I believe that the metrics measured are correlated with safety (risk of collision) - when the user is not aware of their score or how they are being measured. Introducing the score does create some problematic incentives and an unhealthy focus on certain events which may create driver distraction (which I've mentioned earlier). However, I also think it's quite likely that personally I'm driving more safely when complying with the Safety Score framework than I was before. I know my wife is a lot happier in the car now; she doesn't like to fly around the car on corners, and I've never understood why. But I can't fight it.
Good correlation of the safety metrics doesn't mean that it's a good measure for a particular driver. It'll be a good predictor for some drivers, and a bad predictor for others. That's fine.
But not my state. I do believe in complying with traffic laws, but I also don't wish to get rear-ended. My wife has complained about my stopping habits in the past - now I am driving correctly. Just this morning I pushed slightly to get through an intersection that I would have previously stopped for. The light was yellow for my entire transit through the intersection. Now I have to work on being more decisive; it takes some adjustment in my threshold - this morning I very briefly eased off the accelerator before realizing the brake was going to be required (so I reapplied the accelerator). There was no traffic on the side streets, just a left turning truck which triggered the light change. So it was also low risk.
Last edited:
KArnold
Active Member
I have the solution for the Safety Score, and it's simple. Just ask each participant's spouse how safe a driver the participant is on a scale of 0 to 100. We don't need any stinking algorithms! Done and accurate! 
I'm guessing I'd be about a 27 on that scale.
I'm guessing I'd be about a 27 on that scale.
Guesswork is never fact. Kind of the opposite. My whole post is clearly guesswork.
As to hype, I'm far from the only person to suggest Elon hypes/overpromises more than most.
Seems very unlikely, but in a previous tweet, he did say 7 days of driving data would be used. So maybe that would filter out people doing a single drive and getting a 100 score.
But he could've been wrong about that, and maybe they aren't looking at days at all. But i still think anyone excited enough for the FSD beta would've pushed it on day 1
I wonder to what extent if any Elon has discussed Tesla's FSD risk exposure with his lawyers and if that has affected the rollout change?
I bet the lawyers want to talk with Elon
he also mentioned a sample size of 7 days driving initially. Seem like signing up at the last minute would prevent you from gaining access to the first rollout.
I still think I should be first this weekend Have a safety score of 100 and a perfect driving record since 1968 when I got my license. No accidents/dings or moving violations and drive in greater Boston. Can't say the same for my Trek/Cannondale cycling where I'm a bit too aggressive.
Have a safety score of 100 and a perfect driving record since 1968 when I got my license. No accidents/dings or moving violations and drive in greater Boston. Can't say the same for my Trek/Cannondale cycling where I'm a bit too aggressive.It's really interesting that they:
1) Came up with an "safety score" using previous data
2) That score is highly non-linear and biases everyone in the 80-100 range
3) That score weights things in a non-intuitive way
4) A score of 100 represents a collision rate lower than the average Tesla rate, yet no scores above 100 are possible
5) After a week of data, magically there are almost exactly 1000 people in the group.
Hmm, I wonder if they actually wanted something that gave them 1K "safest people" and actually created an algorithm that would to pick ~1K people if run for a week on a the data set, instead of being only focused on relative safety of drivers. For all we know it puts 1K at 100, 2K at 99, 3K at 98, etc, so it's actually just an engineered roll out under the PR of "safety."
1) Came up with an "safety score" using previous data
2) That score is highly non-linear and biases everyone in the 80-100 range
3) That score weights things in a non-intuitive way
4) A score of 100 represents a collision rate lower than the average Tesla rate, yet no scores above 100 are possible
5) After a week of data, magically there are almost exactly 1000 people in the group.
Hmm, I wonder if they actually wanted something that gave them 1K "safest people" and actually created an algorithm that would to pick ~1K people if run for a week on a the data set, instead of being only focused on relative safety of drivers. For all we know it puts 1K at 100, 2K at 99, 3K at 98, etc, so it's actually just an engineered roll out under the PR of "safety."
Just got my score back to 100, I’ll be curious if he is looking for the perfect 100’s or rounded 100’s. I won’t be driving the car tomorrow in fear of knocking the score, we shall see how this turns out.
You think they have some future prediction algorithm that would find exactly 1k people that will get 100/100? For someone who thinks Elon is just flailing around with no plan and no real path to L5 FSD (not saying you're wrong), I think you're giving him too much credit
Last edited:
FalconFour
Member
Guess who's in the club 
Thanks to whoever wrote Safety Score Calculator and the explanation of daily weighted scores, which I eventually figured out just in time for a 99->100 correction.
Thanks to whoever wrote Safety Score Calculator and the explanation of daily weighted scores, which I eventually figured out just in time for a 99->100 correction.
Similar threads
